planner/core: thoroughly push down count-distinct agg in the MPP mode. (#25662)

[ti-srebot]

cherry-pick #25662 to release-5.0
You can switch your code base to this Pull Request by using git-extras:

# In tidb repo:
git pr https://github.com/pingcap/tidb/pull/25861

After apply modifications, you can push your change to this PR via:

git push git@github.com:ti-srebot/tidb.git pr/25861:release-5.0-3ad894da97d9

---

…up by columns

What problem does this PR solve?

Problem Summary:
Right now for count-distinct agg we can only execute the dedup part on TiFlash and gather all the intermediate data to get final result on TiDB. However the execution in TiDB will be the performance bottleneck so we hope to push down this part in MPP mode.

What is changed and how it works?

What's Changed:
We check the group by columns and distinct functions during exhaustPhysicalPlans and set the physical hint to MppScalar indicating this is a scalar function then we will push down all the calculation. This tune only takes effect in the MPP mode.

How it Works:
for a simple statement: select count(distinct a) from t;. The performace advance is great:
before tune:

mysql> explain analyze select count(distinct a) from t;
+------------------------------+-------------+----------+-------------------+---------------+-----------------------------------------------------------------------------------------------------------------+------------------------------------------+----------+------+
| id                           | estRows     | actRows  | task              | access object | execution info                                                                                                  | operator info                            | memory   | disk |
+------------------------------+-------------+----------+-------------------+---------------+-----------------------------------------------------------------------------------------------------------------+------------------------------------------+----------+------+
| HashAgg_13                   | 1.00        | 1        | root              |               | time:18.2s, loops:2                                                                                             | funcs:count(distinct test.t.a)->Column#4 | 932.1 MB | N/A  |
| └─TableReader_15             | 1.00        | 30950159 | root              |               | time:461.5ms, loops:30593, cop_task: {num: 640, max: 0s, min: 0s, avg: 0s, p95: 0s, copr_cache_hit_ratio: 0.00} | data:ExchangeSender_14                   | N/A      | N/A  |
|   └─ExchangeSender_14        | 1.00        | 30950159 | batchCop[tiflash] |               | tiflash_task:{proc max:1.11s, min:0s, p80:1ms, p95:8ms, iters:768, tasks:640, threads:48}                       | ExchangeType: PassThrough                | N/A      | N/A  |
|     └─HashAgg_6              | 1.00        | 30950159 | batchCop[tiflash] |               | tiflash_task:{proc max:1.11s, min:0s, p80:1ms, p95:8ms, iters:768, tasks:640, threads:3}                        | group by:test.t.a,                       | N/A      | N/A  |
|       └─TableFullScan_12     | 64349746.00 | 64238857 | batchCop[tiflash] | table:t       | tiflash_task:{proc max:274.6ms, min:0s, p80:0s, p95:0s, iters:1044, tasks:640, threads:48}                      | keep order:false                         | N/A      | N/A  |
+------------------------------+-------------+----------+-------------------+---------------+-----------------------------------------------------------------------------------------------------------------+------------------------------------------+----------+------+
5 rows in set (18.22 sec)

after tune

mysql> explain analyze select count(distinct a) from t;
+----------------------------------+----------+----------+-------------------+---------------+--------------------------------------------------------------------------------------------+------------------------------------------+--------+------+
| id                               | estRows  | actRows  | task              | access object | execution info                                                                             | operator info                            | memory | disk |
+----------------------------------+----------+----------+-------------------+---------------+--------------------------------------------------------------------------------------------+------------------------------------------+--------+------+
| TableReader_18                   | 1.00     | 1        | root              |               | time:3.29s, loops:2, cop_task: {num: 1, max: 0s, proc_keys: 0, copr_cache_hit_ratio: 0.00} | data:ExchangeSender_17                   | N/A    | N/A  |
| └─ExchangeSender_17              | 1.00     | 1        | batchCop[tiflash] |               | tiflash_task:{time:3.19s, loops:1, threads:20}                                             | ExchangeType: PassThrough                | N/A    | N/A  |
|   └─HashAgg_14                   | 1.00     | 1        | batchCop[tiflash] |               | tiflash_task:{time:3.19s, loops:1, threads:1}                                              | funcs:count(distinct test.t.a)->Column#4 | N/A    | N/A  |
|     └─ExchangeReceiver_16        | 1.00     | 30950159 | batchCop[tiflash] |               | tiflash_task:{time:1.09s, loops:640, threads:20}                                           |                                          | N/A    | N/A  |
|       └─ExchangeSender_15        | 1.00     | 30950159 | batchCop[tiflash] |               | tiflash_task:{time:1.23s, loops:768, threads:48}                                           | ExchangeType: PassThrough                | N/A    | N/A  |
|         └─HashAgg_6              | 1.00     | 30950159 | batchCop[tiflash] |               | tiflash_task:{time:1.23s, loops:768, threads:3}                                            | group by:test.t.a,                       | N/A    | N/A  |
|           └─TableFullScan_12     | 10000.00 | 64238857 | batchCop[tiflash] | table:t       | tiflash_task:{time:256.9ms, loops:1044, threads:48}                                        | keep order:false, stats:pseudo           | N/A    | N/A  |
+----------------------------------+----------+----------+-------------------+---------------+--------------------------------------------------------------------------------------------+------------------------------------------+--------+------+
7 rows in set (6.24 sec)

Check List

Tests

• Unit test
• Manual test (add detailed scripts or steps below)

Release note

• planner/core: thoroughly push down count-distinct agg in the MPP mode.

[ti-srebot]

/run-all-tests

[ti-srebot]

@hanfei1991 you're already a collaborator in bot's repo.

[ti-chi-bot]

[REVIEW NOTIFICATION]

This pull request has been approved by:

• qw4990
• rebelice

To complete the pull request process, please ask the reviewers in the list to review by filling /cc @reviewer in the comment.
After your PR has acquired the required number of LGTMs, you can assign this pull request to the committer in the list by filling /assign @committer in the comment to help you merge this pull request.

The full list of commands accepted by this bot can be found here.

Reviewer can indicate their review by submitting an approval review.
Reviewer can cancel approval by submitting a request changes review.

[zhouqiang-cl]

/merge

[ti-chi-bot]

This pull request has been accepted and is ready to merge.

Commit hash: 3f09286

[hanfei1991]

/merge

[ti-chi-bot]

@ti-srebot: Your PR was out of date, I have automatically updated it for you.

At the same time I will also trigger all tests for you:

/run-all-tests

If the CI test fails, you just re-trigger the test that failed and the bot will merge the PR for you after the CI passes.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the ti-community-infra/tichi repository.