sunchao opened a new pull request, #39633:
URL: https://github.com/apache/spark/pull/39633
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines:
https://spark.apache.org/contributing.html
2. Ensure you have added or run the appropriate tests for your PR:
https://spark.apache.org/developer-tools.html
3. If the PR is unfinished, add '[WIP]' in your PR title, e.g.,
'[WIP][SPARK-XXXX] Your PR title ...'.
4. Be sure to keep the PR description updated to reflect all changes.
5. Please write your PR title to summarize what this PR proposes.
6. If possible, provide a concise example to reproduce the issue for a
faster review.
7. If you want to add a new configuration, please read the guideline first
for naming configurations in
'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.
8. If you want to add or modify an error type or message, please read the
guideline first in
'core/src/main/resources/error/README.md'.
-->
### What changes were proposed in this pull request?
<!--
Please clarify what changes you are proposing. The purpose of this section
is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster
reviews in your PR. See the examples below.
1. If you refactor some codes with changing classes, showing the class
hierarchy will help reviewers.
2. If you fix some SQL features, you can provide some references of other
DBMSes.
3. If there is design documentation, please add the link.
4. If there is a discussion in the mailing list, please add the link.
-->
Currently with [storage-partitioned
join](https://docs.google.com/document/d/1foTkDSM91VxKgkEcBMsuAvEjNybjja-uHk-r3vtXWFE/edit#heading=h.82w8qxfl2uwl),
both sides of the join must be **fully clustered** over the partition values,
that is, each Spark partition should have a distinct partition value. To
guarantee this, Spark will group all the input partitions reported by a V2 data
source on the partition values. The consequence, however, is that it can easily
lead to data skew, when a few partition values have large amount of data.
This PR introduce a new mechanism which requires only one side of the
storage-partitioned join to be fully clustered, while the other side can be
**partially clustered**, i.e., there could exist multiple Spark partitions with
the identical partition value. At planning time, Spark will compare the
statistics from both sides of the join, and pick the side with smaller size as
full clustered side, while the other side is partially clustered. It then
replicate the partitions on the former to match the number of partitions on the
latter.
A new config
`spark.sql.sources.v2.bucketing.partiallyClusteredDistribution.enabled` is
added to enable & disable the feature. By default it is turned off.
For instance, consider the following SQL query:
```sql
SELECT * FROM a JOIN b on a.id = b.id
```
while table `a` reports partitions `[0, 1, 2]` while `b` reports partitions
`[0, 1, 1, 1, 2, 2]`.
Without the PR, Spark would group input partitions so that both sides will
have 3 partitions `[0, 1, 2]`. With the PR, Spark will choose the right-hand
side as partially clustered and match the left-hand side with it. Therefore,
both sides have 6 partitions `[0, 1, 1, 1, 2, 2]`.
Note this PR currently relies on a simple heuristic and always pick the side
with less data size based on table statistics as the side fully clustered, even
though it could also contain skewed partitions. In future, we can potentially
do fine-grained comparison based on partition values.
### Why are the changes needed?
<!--
Please clarify why the changes are needed. For instance,
1. If you propose a new API, clarify the use case for a new API.
2. If you fix a bug, you can clarify why it is a bug.
-->
As mentioned in the previous section, this feature can help to reduce data
skewness when a partition value is mapped to a large amount of rows.
### Does this PR introduce _any_ user-facing change?
<!--
Note that it means *any* user-facing change including all aspects such as
the documentation fix.
If yes, please clarify the previous behavior and the change this PR proposes
- provide the console output, description and/or an example to show the
behavior difference if possible.
If possible, please also clarify if this is a user-facing change compared to
the released Spark versions or within the unreleased branches such as master.
If no, write 'No'.
-->
A new Spark config,
`spark.sql.sources.v2.bucketing.partiallyClusteredDistribution.enabled`, is
introduced to enable/disable the feature. By default it is disabled, so the
behavior should still be the same as before.
### How was this patch tested?
<!--
If tests were added, say they were added here. Please make sure to add some
test cases that check the changes thoroughly including negative and positive
cases if possible.
If it was tested in a way different from regular unit tests, please clarify
how you tested step by step, ideally copy and paste-able, so that other
reviewers can test and check, and descendants can verify in the future.
If tests were not added, please describe why they were not added and/or why
it was difficult to add.
If benchmark tests were added, please run the benchmarks in GitHub Actions
for the consistent environment, and the instructions could accord to:
https://spark.apache.org/developer-tools.html#github-workflow-benchmarks.
-->
Added a few new tests in `KeyGroupedPartitioningSuite`.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
