ODIn/spark-instrumented-optimizer
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity
spark-instrumented-optimizer/sql/core/benchmarks/DateTimeBenchmark-jdk11-results.txt

430 lines
35 KiB
Plaintext
Raw Normal View History

[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Extract components
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
cast to timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
cast to timestamp wholestage off 546 572 36 18.3 54.6 1.0X
cast to timestamp wholestage on 412 438 16 24.3 41.2 1.3X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
year of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
year of timestamp wholestage off 1240 1295 77 8.1 124.0 1.0X
year of timestamp wholestage on 1109 1130 24 9.0 110.9 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
quarter of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
quarter of timestamp wholestage off 1572 1574 3 6.4 157.2 1.0X
quarter of timestamp wholestage on 1386 1405 18 7.2 138.6 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
month of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
month of timestamp wholestage off 1194 1196 2 8.4 119.4 1.0X
month of timestamp wholestage on 1057 1069 12 9.5 105.7 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
weekofyear of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
weekofyear of timestamp wholestage off 2070 2071 2 4.8 207.0 1.0X
weekofyear of timestamp wholestage on 1549 1555 6 6.5 154.9 1.3X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
day of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
day of timestamp wholestage off 1173 1186 18 8.5 117.3 1.0X
day of timestamp wholestage on 1056 1076 26 9.5 105.6 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
dayofyear of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
dayofyear of timestamp wholestage off 1207 1211 6 8.3 120.7 1.0X
dayofyear of timestamp wholestage on 1097 1108 9 9.1 109.7 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
dayofmonth of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
dayofmonth of timestamp wholestage off 1184 1190 8 8.4 118.4 1.0X
dayofmonth of timestamp wholestage on 1053 1060 9 9.5 105.3 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
dayofweek of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
dayofweek of timestamp wholestage off 1343 1362 27 7.4 134.3 1.0X
dayofweek of timestamp wholestage on 1228 1239 7 8.1 122.8 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
weekday of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
weekday of timestamp wholestage off 1276 1278 3 7.8 127.6 1.0X
weekday of timestamp wholestage on 1160 1181 22 8.6 116.0 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
hour of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
hour of timestamp wholestage off 854 862 12 11.7 85.4 1.0X
hour of timestamp wholestage on 741 748 6 13.5 74.1 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
minute of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
minute of timestamp wholestage off 853 854 1 11.7 85.3 1.0X
minute of timestamp wholestage on 730 737 11 13.7 73.0 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
second of timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
second of timestamp wholestage off 726 728 2 13.8 72.6 1.0X
second of timestamp wholestage on 614 623 9 16.3 61.4 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Current date and time
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
current_date: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
current_date wholestage off 369 370 2 27.1 36.9 1.0X
current_date wholestage on 277 284 8 36.2 27.7 1.3X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
current_timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
current_timestamp wholestage off 410 411 1 24.4 41.0 1.0X
current_timestamp wholestage on 283 418 259 35.4 28.3 1.5X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Date arithmetic
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
cast to date: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
cast to date wholestage off 987 992 7 10.1 98.7 1.0X
cast to date wholestage on 891 896 4 11.2 89.1 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
last_day: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
last_day wholestage off 1179 1179 1 8.5 117.9 1.0X
last_day wholestage on 1052 1080 44 9.5 105.2 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
next_day: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
next_day wholestage off 1073 1081 11 9.3 107.3 1.0X
next_day wholestage on 948 955 10 10.5 94.8 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_add: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_add wholestage off 1006 1009 5 9.9 100.6 1.0X
date_add wholestage on 867 870 4 11.5 86.7 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_sub: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_sub wholestage off 980 988 11 10.2 98.0 1.0X
date_sub wholestage on 866 873 9 11.6 86.6 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
add_months: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
add_months wholestage off 1329 1332 4 7.5 132.9 1.0X
add_months wholestage on 1199 1206 8 8.3 119.9 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Formatting dates
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
format date: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
format date wholestage off 4724 4736 18 2.1 472.4 1.0X
format date wholestage on 4550 4574 26 2.2 455.0 1.0X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Formatting timestamps
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
from_unixtime: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
from_unixtime wholestage off 7171 7183 17 1.4 717.1 1.0X
from_unixtime wholestage on 7114 7141 20 1.4 711.4 1.0X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Convert timestamps
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
from_utc_timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
from_utc_timestamp wholestage off 1498 1504 8 6.7 149.8 1.0X
from_utc_timestamp wholestage on 1399 1405 5 7.1 139.9 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
to_utc_timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
to_utc_timestamp wholestage off 1505 1507 2 6.6 150.5 1.0X
to_utc_timestamp wholestage on 1396 1401 5 7.2 139.6 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Intervals
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
cast interval: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
cast interval wholestage off 423 428 7 23.6 42.3 1.0X
cast interval wholestage on 300 302 2 33.3 30.0 1.4X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
datediff: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
datediff wholestage off 1626 1630 6 6.1 162.6 1.0X
datediff wholestage on 1467 1471 3 6.8 146.7 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
months_between: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
months_between wholestage off 1988 1992 5 5.0 198.8 1.0X
months_between wholestage on 1812 1834 24 5.5 181.2 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
window: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
window wholestage off 2277 2334 80 0.4 2277.1 1.0X
window wholestage on 48996 49048 67 0.0 48996.0 0.0X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Truncation
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc YEAR: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc YEAR wholestage off 867 870 6 11.5 86.7 1.0X
date_trunc YEAR wholestage on 815 819 6 12.3 81.5 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc YYYY: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc YYYY wholestage off 866 875 13 11.5 86.6 1.0X
date_trunc YYYY wholestage on 811 813 2 12.3 81.1 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc YY: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc YY wholestage off 864 867 4 11.6 86.4 1.0X
date_trunc YY wholestage on 812 824 10 12.3 81.2 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc MON: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc MON wholestage off 881 884 4 11.3 88.1 1.0X
date_trunc MON wholestage on 820 826 7 12.2 82.0 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc MONTH: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc MONTH wholestage off 880 881 2 11.4 88.0 1.0X
date_trunc MONTH wholestage on 819 822 4 12.2 81.9 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc MM: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc MM wholestage off 889 904 21 11.2 88.9 1.0X
date_trunc MM wholestage on 818 828 8 12.2 81.8 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc DAY: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc DAY wholestage off 590 593 4 16.9 59.0 1.0X
date_trunc DAY wholestage on 510 514 4 19.6 51.0 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc DD: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc DD wholestage off 596 604 11 16.8 59.6 1.0X
date_trunc DD wholestage on 511 519 9 19.6 51.1 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc HOUR: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc HOUR wholestage off 586 592 9 17.1 58.6 1.0X
date_trunc HOUR wholestage on 507 513 5 19.7 50.7 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc MINUTE: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc MINUTE wholestage off 561 562 1 17.8 56.1 1.0X
date_trunc MINUTE wholestage on 480 485 4 20.8 48.0 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc SECOND: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc SECOND wholestage off 561 561 1 17.8 56.1 1.0X
date_trunc SECOND wholestage on 479 480 2 20.9 47.9 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc WEEK: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc WEEK wholestage off 725 727 2 13.8 72.5 1.0X
date_trunc WEEK wholestage on 674 684 11 14.8 67.4 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
date_trunc QUARTER: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
date_trunc QUARTER wholestage off 1653 1659 10 6.1 165.3 1.0X
date_trunc QUARTER wholestage on 1588 1601 12 6.3 158.8 1.0X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
trunc year: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
trunc year wholestage off 391 393 2 25.6 39.1 1.0X
trunc year wholestage on 312 316 4 32.1 31.2 1.3X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
trunc yyyy: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
trunc yyyy wholestage off 390 394 5 25.6 39.0 1.0X
trunc yyyy wholestage on 316 319 4 31.6 31.6 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
trunc yy: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
trunc yy wholestage off 387 390 5 25.8 38.7 1.0X
trunc yy wholestage on 313 316 3 31.9 31.3 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
trunc mon: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
trunc mon wholestage off 388 395 11 25.8 38.8 1.0X
trunc mon wholestage on 314 316 2 31.8 31.4 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
trunc month: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
trunc month wholestage off 388 390 4 25.8 38.8 1.0X
trunc month wholestage on 315 318 2 31.7 31.5 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
trunc mm: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
trunc mm wholestage off 384 388 4 26.0 38.4 1.0X
trunc mm wholestage on 314 321 9 31.9 31.4 1.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Parsing
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
to timestamp str: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
to timestamp str wholestage off 178 191 18 5.6 178.5 1.0X
to timestamp str wholestage on 157 160 2 6.4 156.6 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
to_timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
to_timestamp wholestage off 1790 1795 7 0.6 1790.0 1.0X
to_timestamp wholestage on 1813 1820 10 0.6 1813.1 1.0X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
to_unix_timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
to_unix_timestamp wholestage off 1836 1837 1 0.5 1835.8 1.0X
to_unix_timestamp wholestage on 1786 1791 3 0.6 1785.8 1.0X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
to date str: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
to date str wholestage off 169 169 1 5.9 168.8 1.0X
to date str wholestage on 151 153 2 6.6 151.2 1.1X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
to_date: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
to_date wholestage off 2504 2512 10 0.4 2504.4 1.0X
to_date wholestage on 2522 2536 19 0.4 2522.3 1.0X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
================================================================================================
Conversion from/to external types
================================================================================================
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
OpenJDK 64-Bit Server VM 11.0.5+10-post-Ubuntu-0ubuntu1.118.04 on Linux 4.15.0-1044-aws
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
To/from java.sql.Timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
[SPARK-30409][SPARK-29173][SQL][TESTS] Use `NoOp` datasource in SQL benchmarks ### What changes were proposed in this pull request? In the PR, I propose to replace `.collect()`, `.count()` and `.foreach(_ => ())` in SQL benchmarks and use the `NoOp` datasource. I added an implicit class to `SqlBasedBenchmark` with the `.noop()` method. It can be used in benchmark like: `ds.noop()`. The last one is unfolded to `ds.write.format("noop").mode(Overwrite).save()`. ### Why are the changes needed? To avoid additional overhead that `collect()` (and other actions) has. For example, `.collect()` has to convert values according to external types and pull data to the driver. This can hide actual performance regressions or improvements of benchmarked operations. ### Does this PR introduce any user-facing change? No ### How was this patch tested? Re-run all modified benchmarks using Amazon EC2. | Item | Description | | ---- | ----| | Region | us-west-2 (Oregon) | | Instance | r3.xlarge (spot instance) | | AMI | ami-06f2f779464715dc5 (ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1) | | Java | OpenJDK8/10 | - Run `TPCDSQueryBenchmark` using instructions from the PR #26049 ``` # `spark-tpcds-datagen` needs this. (JDK8) $ git clone https://github.com/apache/spark.git -b branch-2.4 --depth 1 spark-2.4 $ export SPARK_HOME=$PWD $ ./build/mvn clean package -DskipTests # Generate data. (JDK8) $ git clone gitgithub.com:maropu/spark-tpcds-datagen.git $ cd spark-tpcds-datagen/ $ build/mvn clean package $ mkdir -p /data/tpcds $ ./bin/dsdgen --output-location /data/tpcds/s1 // This need `Spark 2.4` ``` - Other benchmarks ran by the script: ``` #!/usr/bin/env python3 import os from sparktestsupport.shellutils import run_cmd benchmarks = [ ['sql/test', 'org.apache.spark.sql.execution.benchmark.AggregateBenchmark'], ['avro/test', 'org.apache.spark.sql.execution.benchmark.AvroReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.BloomFilterBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DataSourceReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ExtractBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.FilterPushdownBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.InExpressionBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.IntervalBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.JoinBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MakeDateTimeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.MiscBenchmark'], ['hive/test', 'org.apache.spark.sql.execution.benchmark.ObjectHashAggregateExecBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.OrcV2NestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.ParquetNestedSchemaPruningBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.RangeBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.UDFBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideSchemaBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.benchmark.WideTableBenchmark'], ['hive/test', 'org.apache.spark.sql.hive.orc.OrcReadBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'], ['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark'] ] print('Set SPARK_GENERATE_BENCHMARK_FILES=1') os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1' for b in benchmarks: print("Run benchmark: %s" % b[1]) run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])]) ``` Closes #27078 from MaxGekk/noop-in-benchmarks. Lead-authored-by: Maxim Gekk <max.gekk@gmail.com> Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com> Co-authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2020-01-12 16:18:19 -05:00
From java.sql.Timestamp 346 367 35 14.5 69.2 1.0X
Collect longs 2139 2329 289 2.3 427.7 0.2X
Collect timestamps 1883 2086 303 2.7 376.5 0.2X
[SPARK-29320][TESTS] Compare `sql/core` module in JDK8/11 (Part 1) ### What changes were proposed in this pull request? This PR regenerates the `sql/core` benchmarks in JDK8/11 to compare the result. In general, we compare the ratio instead of the time. However, in this PR, the average time is compared. This PR should be considered as a rough comparison. **A. EXPECTED CASES(JDK11 is faster in general)** - [x] BloomFilterBenchmark (JDK11 is faster except one case) - [x] BuiltInDataSourceWriteBenchmark (JDK11 is faster at CSV/ORC) - [x] CSVBenchmark (JDK11 is faster except five cases) - [x] ColumnarBatchBenchmark (JDK11 is faster at `boolean`/`string` and some cases in `int`/`array`) - [x] DatasetBenchmark (JDK11 is faster with `string`, but is slower for `long` type) - [x] ExternalAppendOnlyUnsafeRowArrayBenchmark (JDK11 is faster except two cases) - [x] ExtractBenchmark (JDK11 is faster except HOUR/MINUTE/SECOND/MILLISECONDS/MICROSECONDS) - [x] HashedRelationMetricsBenchmark (JDK11 is faster) - [x] JSONBenchmark (JDK11 is much faster except eight cases) - [x] JoinBenchmark (JDK11 is faster except five cases) - [x] OrcNestedSchemaPruningBenchmark (JDK11 is faster in nine cases) - [x] PrimitiveArrayBenchmark (JDK11 is faster) - [x] SortBenchmark (JDK11 is faster except `Arrays.sort` case) - [x] UDFBenchmark (N/A, values are too small) - [x] UnsafeArrayDataBenchmark (JDK11 is faster except one case) - [x] WideTableBenchmark (JDK11 is faster except two cases) **B. CASES WE NEED TO INVESTIGATE MORE LATER** - [x] AggregateBenchmark (JDK11 is slower in general) - [x] CompressionSchemeBenchmark (JDK11 is slower in general except `string`) - [x] DataSourceReadBenchmark (JDK11 is slower in general) - [x] DateTimeBenchmark (JDK11 is slightly slower in general except `parsing`) - [x] MakeDateTimeBenchmark (JDK11 is slower except two cases) - [x] MiscBenchmark (JDK11 is slower except ten cases) - [x] OrcV2NestedSchemaPruningBenchmark (JDK11 is slower) - [x] ParquetNestedSchemaPruningBenchmark (JDK11 is slower except six cases) - [x] RangeBenchmark (JDK11 is slower except one case) `FilterPushdownBenchmark/InExpressionBenchmark/WideSchemaBenchmark` will be compared later because it took long timer. ### Why are the changes needed? According to the result, there are some difference between JDK8/JDK11. This will be a baseline for the future improvement and comparison. Also, as a reproducible environment, the following environment is used. - Instance: `r3.xlarge` - OS: `CentOS Linux release 7.5.1804 (Core)` - JDK: - `OpenJDK Runtime Environment (build 1.8.0_222-b10)` - `OpenJDK Runtime Environment 18.9 (build 11.0.4+11-LTS)` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? This is a test-only PR. We need to run benchmark. Closes #26003 from dongjoon-hyun/SPARK-29320. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
2019-10-03 11:58:25 -04:00
Reference in a new issue Copy permalink