[Enhancement] optimize the performace for topn with large offset #55886
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
![starrocks-cr[bot]](https://avatars.githubusercontent.com/in/404334?s=60&v=4){kind=small}
stdpain
force-pushed
the
opt_topn_with_large_limit
branch
from
8eb9d93 to
2bf3116
Compare
satanson
previously approved these changes
Feb 14, 2025
stdpain
force-pushed
the
opt_topn_with_large_limit
branch
2 times, most recently
from
6735781 to
d421e44
Compare
satanson
previously approved these changes
Feb 24, 2025
stdpain
force-pushed
the
opt_topn_with_large_limit
branch
4 times, most recently
from
160e90d to
e765fd7
Compare
Signed-off-by: stdpain <[email protected]>
stdpain
force-pushed
the
opt_topn_with_large_limit
branch
from
e765fd7 to
4262201
Compare
[FE Incremental Coverage Report]✅ pass : 0 / 0 (0%) |
[Java-Extensions Incremental Coverage Report]✅ pass : 0 / 0 (0%) |
[BE Incremental Coverage Report]✅ pass : 296 / 344 (86.05%) file detail
|
murphyatwork
approved these changes
Feb 26, 2025
Seaven
approved these changes
Feb 27, 2025
|
@Mergifyio backport branch-3.3 |
✅ Backports have been created
|
|
@Mergifyio backport branch-3.4 |
✅ Backports have been created
|
mergify bot
pushed a commit
that referenced
this pull request
Feb 27, 2025
) 1. change the max_buffered_size to chunk_size/4096 when limit greater than 65535 2. Avoid large permutations that take up too much memory. 3. Reduce memory when merging large chunks SSB100G dop=4 1BE ``` select lo_shipmode from lineorder order by lo_shipmode limit 50000000, 400 ``` baseline:1m42s patched:28s339ms The reasons why baseline performance is too low are: 1.The merge operation is too frequent and needs to be done every 256 chunks. But the total input data is too large. This results in too many merge operations. This PR adds max_buffer_size, which depends on offset + limit /chunk_size, to reduce the frequency of merge operations. But it may result in using more memory. This PR additionally optimizes memory for merge chunks. It can reduce the peak memory of merge. Signed-off-by: stdpain <[email protected]> (cherry picked from commit 11b98b0) # Conflicts: # be/src/exec/chunks_sorter.cpp # be/src/exec/chunks_sorter.h # be/src/exec/chunks_sorter_topn.h # be/src/exec/pipeline/sort/local_partition_topn_context.cpp # be/src/exec/sorting/merge.h
18 tasks
mergify bot
pushed a commit
that referenced
this pull request
Feb 27, 2025
) 1. change the max_buffered_size to chunk_size/4096 when limit greater than 65535 2. Avoid large permutations that take up too much memory. 3. Reduce memory when merging large chunks SSB100G dop=4 1BE ``` select lo_shipmode from lineorder order by lo_shipmode limit 50000000, 400 ``` baseline:1m42s patched:28s339ms The reasons why baseline performance is too low are: 1.The merge operation is too frequent and needs to be done every 256 chunks. But the total input data is too large. This results in too many merge operations. This PR adds max_buffer_size, which depends on offset + limit /chunk_size, to reduce the frequency of merge operations. But it may result in using more memory. This PR additionally optimizes memory for merge chunks. It can reduce the peak memory of merge. Signed-off-by: stdpain <[email protected]> (cherry picked from commit 11b98b0)
18 tasks
wanpengfei-git
pushed a commit
that referenced
this pull request
Feb 27, 2025
…kport #55886) (#56370) Co-authored-by: stdpain <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Why I'm doing:
SSB100G dop=4 1BE
baseline:1m42s patched:28s339ms
The reasons why baseline performance is too low are:
1.The merge operation is too frequent and needs to be done every 256 chunks. But the total input data is too large. This results in too many merge operations.
This PR adds max_buffer_size, which depends on offset + limit /chunk_size, to reduce the frequency of merge operations. But it may result in using more memory.
This PR additionally optimizes memory for merge chunks. It can reduce the peak memory of merge.
What I'm doing:
Fixes #issue
What type of PR is this:
Does this PR entail a change in behavior?
If yes, please specify the type of change:
Checklist:
Bugfix cherry-pick branch check: