ODIn/spark-instrumented-optimizer
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity
spark-instrumented-optimizer/sql
History
Kazuaki Ishizaki c09b31eb8f [SPARK-21217][SQL] Support ColumnVector.Array.to<type>Array() 
## What changes were proposed in this pull request?

This PR implements bulk-copy for `ColumnVector.Array.to<type>Array()` methods (e.g. `toIntArray()`) in `ColumnVector.Array` by using `System.arrayCopy()` or `Platform.copyMemory()`.

Before this PR, when one of these method is called, the generic method in `ArrayData` is called. It is not fast since element-wise copy is performed.

This PR can improve performance of a benchmark program by 1.9x and 3.2x.

Without this PR
```
OpenJDK 64-Bit Server VM 1.8.0_131-8u131-b11-0ubuntu1.16.04.2-b11 on Linux 4.4.0-66-generic
Intel(R) Xeon(R) CPU E5-2667 v3  3.20GHz

Int Array                                Best/Avg Time(ms)    Rate(M/s)   Per Row(ns)
------------------------------------------------------------------------------------------------
ON_HEAP                                        586 /  628         14.3          69.9
OFF_HEAP                                       893 /  902          9.4         106.5
```

With this PR
```
OpenJDK 64-Bit Server VM 1.8.0_131-8u131-b11-0ubuntu1.16.04.2-b11 on Linux 4.4.0-66-generic
Intel(R) Xeon(R) CPU E5-2667 v3  3.20GHz

Int Array                                Best/Avg Time(ms)    Rate(M/s)   Per Row(ns)
------------------------------------------------------------------------------------------------
ON_HEAP                                        306 /  331         27.4          36.4
OFF_HEAP                                       282 /  287         29.8          33.6
```

Source program
```
    (MemoryMode.ON_HEAP :: MemoryMode.OFF_HEAP :: Nil).foreach { memMode => {
      val len = 8 * 1024 * 1024
      val column = ColumnVector.allocate(len * 2, new ArrayType(IntegerType, false), memMode)

      val data = column.arrayData
      var i = 0
      while (i < len) {
        data.putInt(i, i)
        i += 1
      }
      column.putArray(0, 0, len)

      val benchmark = new Benchmark("Int Array", len, minNumIters = 20)
      benchmark.addCase(s"$memMode") { iter =>
        var i = 0
        while (i < 50) {
          column.getArray(0).toIntArray
          i += 1
        }
      }
      benchmark.run
    }}
```

## How was this patch tested?

Added test suite

Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>

Closes #18425 from kiszk/SPARK-21217.
2017-07-07 13:09:32 +08:00
..
catalyst [SPARK-21323][SQL] Rename plans.logical.statsEstimation.Range to ValueInterval 2017-07-06 13:58:27 -07:00
core [SPARK-21217][SQL] Support ColumnVector.Array.to<type>Array() 2017-07-07 13:09:32 +08:00
hive [SPARK-21324][TEST] Improve statistics test suites 2017-07-06 16:00:31 +08:00
hive-thriftserver [SPARK-20345][SQL] Fix STS error handling logic on HiveSQLException 2017-06-12 14:05:03 -07:00
README.md [SPARK-16557][SQL] Remove stale doc in sql/README.md 2016-07-14 19:24:42 -07:00

README.md

Spark SQL

This module provides support for executing relational queries expressed in either SQL or the DataFrame/Dataset API.

Spark SQL is broken up into four subprojects:

  • Catalyst (sql/catalyst) - An implementation-agnostic framework for manipulating trees of relational operators and expressions.
  • Execution (sql/core) - A query planner / execution engine for translating Catalyst's logical query plans into Spark RDDs. This component also includes a new public interface, SQLContext, that allows users to execute SQL or LINQ statements against existing RDDs and Parquet files.
  • Hive Support (sql/hive) - Includes an extension of SQLContext called HiveContext that allows users to write queries using a subset of HiveQL and access data from a Hive Metastore using Hive SerDes. There are also wrappers that allows users to run queries that include Hive UDFs, UDAFs, and UDTFs.
  • HiveServer and CLI support (sql/hive-thriftserver) - Includes support for the SQL CLI (bin/spark-sql) and a HiveServer2 (for JDBC/ODBC) compatible server.