Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

storage: Abort vector index building as soon as possible #9443

Conversation

Lloyd-Pottiger
Copy link
Contributor

@Lloyd-Pottiger Lloyd-Pottiger commented Sep 20, 2024

What problem does this PR solve?

Issue Number: close #9032

Problem Summary:

What is changed and how it works?

Pick https://github.com/tidbcloud/tiflash-cse/pull/232, https://github.com/tidbcloud/tiflash-cse/pull/240 and https://github.com/tidbcloud/tiflash-cse/pull/270

  • Use a lower efSearch to possibly improve performance and trade off with recall. Performance can be improved when searching top K where K < 64.
  • Abort vector index building as soon as possible. This could be helpful during the process of loading a lot of vector data.

TODO: In this PR, index build is cancelled after segment stable is invalid, e.g. after applyMergeDelta. Maybe we can do it as soon as possible, e.g. when prepareMergeDelta?

Benchmark @ M1 Pro:

efSearch=64, TopK=10: 107826 ns
efSearch=16, TopK=10: 63858 ns

storage: Abort vector index building as soon as possible

Check List

Tests

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

Run VectorDBBench, and we can see that during the data loading process actually all vector index build will be cancelled:

[2024/07/07 10:47:52.290 +08:00] [DEBUG] [Segment.cpp:1244] ["MergeDelta - Begin apply"] [source="keyspace=1 table_id=96 segment_id=1 epoch=3"] [thread_id=41]
[2024/07/07 10:47:52.291 +08:00] [DEBUG] [Segment.cpp:1277] ["MergeDelta - Finish apply, old_me=<segment_id=1 epoch=3 range=[-9223372036854775808,9223372036854775807) next_segment_id=0 delta_rows=94500 delta_bytes=3591000 delta_deletes=0 stable_file=dmf_4(v=0) stable_rows=220500 stable_bytes=684873000 dmf_rows=220500 dmf_bytes=684873000 dmf_packs=27> new_me=<segment_id=1 epoch=4 range=[-9223372036854775808,9223372036854775807) next_segment_id=0 delta_rows=10500 delta_bytes=399000 delta_deletes=0 stable_file=dmf_5(v=0) stable_rows=304500 stable_bytes=945777000 dmf_rows=304500 dmf_bytes=945777000 dmf_packs=38>"] [source="keyspace=1 table_id=96 segment_id=1 epoch=3"] [thread_id=41]
[2024/07/07 10:47:52.291 +08:00] [DEBUG] [Segment.h:585] ["Abandon segment, segment=<segment_id=1 epoch=3 stable=dmf_4(v=0) range=[-9223372036854775808,9223372036854775807)>"] [source="keyspace=1 table_id=96 segment_id=1 epoch=3"] [thread_id=41]
[2024/07/07 10:47:52.291 +08:00] [INFO] [DeltaMergeStore_InternalSegment.cpp:909] ["MergeDelta - Finish, delta is merged, old_segment=<segment_id=1 epoch=3 range=[-9223372036854775808,9223372036854775807) abandoned=true next_segment_id=0 delta_rows=94500 delta_bytes=3591000 delta_deletes=0 stable_file=dmf_4(v=0) stable_rows=220500 stable_bytes=684873000 dmf_rows=220500 dmf_bytes=684873000 dmf_packs=27> new_segment=<segment_id=1 epoch=4 range=[-9223372036854775808,9223372036854775807) next_segment_id=0 delta_rows=10500 delta_bytes=399000 delta_deletes=0 stable_file=dmf_5(v=0) stable_rows=304500 stable_bytes=945777000 dmf_rows=304500 dmf_bytes=945777000 dmf_packs=38>"] [source="keyspace=1 table_id=96"] [thread_id=41]
...
...
[2024/07/07 10:47:52.478 +08:00] [DEBUG] [DMFileV3IncrementWriter.cpp:125] ["Abandon increment write, local_path=/Users/breezewish/.tiup/data/serverless/tiflash-0/main_data/data/ks_1_t_96/stable/dmf_4 file_names={}"] [thread_id=36]
[2024/07/07 10:47:52.481 +08:00] [INFO] [DeltaMergeStore_InternalSegment.cpp:739] ["EnsureStableIndex - Build index aborted because DMFile is no longer valid, dm_files=dmf_4(v=0) source_segment=<segment_id=1 epoch=3 stable=dmf_4(v=0) range=[-9223372036854775808,9223372036854775807)>"] [source="keyspace=1 table_id=96"] [thread_id=36]
  • No code

Side effects

  • Performance regression: Consumes more CPU
  • Performance regression: Consumes more Memory
  • Breaking backward compatibility

Documentation

  • Affects user behaviors
  • Contains syntax changes
  • Contains variable changes
  • Contains experimental features
  • Changes MySQL compatibility

Release note

None

Signed-off-by: Wish <[email protected]>
Co-authored-by: Lloyd-Pottiger <[email protected]>
@ti-chi-bot ti-chi-bot bot added release-note-none Denotes a PR that doesn't merit a release note. size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. labels Sep 20, 2024
Copy link
Contributor

@JaySon-Huang JaySon-Huang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@ti-chi-bot ti-chi-bot bot added needs-1-more-lgtm Indicates a PR needs 1 more LGTM. approved labels Sep 20, 2024
Copy link
Contributor

ti-chi-bot bot commented Sep 20, 2024

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: breezewish, JaySon-Huang

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

The pull request process is described here

Needs approval from an approver in each of these files:
  • OWNERS [JaySon-Huang,breezewish]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@ti-chi-bot ti-chi-bot bot added lgtm and removed needs-1-more-lgtm Indicates a PR needs 1 more LGTM. labels Sep 20, 2024
Copy link
Contributor

ti-chi-bot bot commented Sep 20, 2024

[LGTM Timeline notifier]

Timeline:

  • 2024-09-20 06:19:23.700022594 +0000 UTC m=+1201233.440446534: ☑️ agreed by JaySon-Huang.
  • 2024-09-20 06:36:56.355795748 +0000 UTC m=+1202286.096219672: ☑️ agreed by breezewish.

@ti-chi-bot ti-chi-bot bot merged commit 4df924b into pingcap:feature/vector-index Sep 20, 2024
5 checks passed
@Lloyd-Pottiger Lloyd-Pottiger deleted the abort-vector-index-build branch September 20, 2024 11:30
@JaySon-Huang JaySon-Huang mentioned this pull request Sep 30, 2024
12 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
approved lgtm release-note-none Denotes a PR that doesn't merit a release note. size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants