Option to change groupers count in ConcurrencyGrouper to avoid contention

[takaaki7]

Description

For my workload, ConcurrencyGrouper's synchronized (subGrouper) { causes many contention and becomes bottleneck.
https://github.com/takaaki7/druid/blob/37853f8de4ba801a5f6c429ccceeb58b58762109/processing/src/main/java/org/apache/druid/query/groupby/epinephelinae/ConcurrentGrouper.java#L290

It seems just increasing groupers can reduce contentions and improve the performance.

Test description

Query

SELECT COUNT(*) AS (SELECT user_id, COUNT(*) AS cnt
FROM event_log
)
WHERE cnt > 100

Input: 600M rows
Intermediate grouping result rows: 20M rows(not filtered by WHERE clause)
Threads: 23 (24core machine)
Latency: 3-4sec

I debugged with jstack and many times threads are stopped at synchronized (subGrouper) { line.

[gianm]

Does it help if you set mergeThreadLocal: true? This removes the synchronization completely.

[takaaki7]

Now I tested and here is the result.
#15227 (comment)

mergeThreadLocal is faster, but the aggregation result is wrong.

[github-actions]

This issue has been marked as stale due to 280 days of inactivity.
It will be closed in 4 weeks if no further activity occurs. If this issue is still
relevant, please simply write any comment. Even if closed, you can still revive the
issue at any time or discuss it on the dev@druid.apache.org list.
Thank you for your contributions.

[github-actions]

This issue has been closed due to lack of activity. If you think that
is incorrect, or the issue requires additional review, you can revive the issue at
any time.