121883b1a5
### What changes were proposed in this pull request? This PR proposes to optimize WAL commit phase via following changes: * cache offset log to avoid FS get operation per batch * just directly delete instead of employing FS list operation on purge ### Why are the changes needed? There're inefficiency on WAL commit phase which can be easily optimized via using a small driver memory. 1. To provide the offset metadata to source side (via `source.commit()`), we read offset metadata for previous batch from file system, which is probably written by this driver in previous batches. Caching it into driver memory would reduce the get operation. 2. Spark calls purge against offset log & commit log per batch, which calls list operation. If the previous batch succeeded to purge, the current batch just needs to check one batch which can be simply done via direct delete operation, instead of calling list operation. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Manually tested with additional debug log. (Verified that cache is used, cache keeps the size as 2, only one delete call is used instead of list call) Did some experiment with simple rate to console query. (NOTE: wasn't done with master branch - tested against Spark 2.4.x, but WAL commit phase hasn't been changed AFAIK during these versions) AWS S3 + S3 guard: > before the patch <img width="1075" alt="aws-before" src="https://user-images.githubusercontent.com/1317309/107108721-6cc54380-687d-11eb-8f10-b906b9d58397.png"> > after the patch <img width="1071" alt="aws-after" src="https://user-images.githubusercontent.com/1317309/107108724-7189f780-687d-11eb-88da-26912ac15c85.png"> Azure: > before the patch <img width="1074" alt="azure-before" src="https://user-images.githubusercontent.com/1317309/107108726-75b61500-687d-11eb-8c06-9048fa10ff9a.png"> > after the patch <img width="1069" alt="azure-after" src="https://user-images.githubusercontent.com/1317309/107108729-79e23280-687d-11eb-8d97-e7f3aeec51be.png"> Closes #31495 from HeartSaVioR/SPARK-34383. Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com> Signed-off-by: Gabor Somogyi <gabor.g.somogyi@gmail.com> |
||
|---|---|---|
| .. | ||
| catalyst | ||
| core | ||
| hive | ||
| hive-thriftserver | ||
| create-docs.sh | ||
| gen-sql-api-docs.py | ||
| gen-sql-config-docs.py | ||
| gen-sql-functions-docs.py | ||
| mkdocs.yml | ||
| 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 extensions that allow users to write queries using a subset of HiveQL and access data from a Hive Metastore using Hive SerDes. There are also wrappers that allow 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.
Running ./sql/create-docs.sh generates SQL documentation for built-in functions under sql/site, and SQL configuration documentation that gets included as part of configuration.md in the main docs directory.