TsReaper opened a new pull request #13636:
URL: https://github.com/apache/flink/pull/13636
## What is the purpose of the change
The current implementation of `HiveTableSource#createBatchSource` for
calculating parallelism directly uses `inputFormat.createInputSplits(0).length`
as the number of splits. However `createInputSplits` may be costly as it will
read some data from all source files, especially when the table is not
partitioned and the number of files are large.
Many Hive tables maintain the number of files in that table, and it's
obvious that the number of splits is at least the number of files. So we can
try to fetch the number of files (almost without cost) first and if the number
of files already exceeds maximum parallelism we can directly use the maximum
parallelism without calling `createInputSplits`.
This is a significant optimization on the current Flink TPCDS benchmark,
which will create some table with 15000 files without partitioning. This
optimization will improve the performance of the whole benchmark by 300s and
more.
## Brief change log
- Optimize parallelism calculating of `HiveTableSource` by checking file
number
## Verifying this change
This change added tests and can be verified as follows: Run newly added test
case in `HiveTableSourceITCase`.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): no
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: no
- The serializers: no
- The runtime per-record code paths (performance sensitive): no
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Kubernetes/Yarn/Mesos, ZooKeeper: no
- The S3 file system connector: no
## Documentation
- Does this pull request introduce a new feature? no
- If yes, how is the feature documented? not applicable
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
