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

Limit segment size for better indexing performance #135

Closed
generall opened this issue Dec 2, 2021 · 1 comment
Closed

Limit segment size for better indexing performance #135

generall opened this issue Dec 2, 2021 · 1 comment

Comments

@generall
Copy link
Member

generall commented Dec 2, 2021

Is your feature request related to a problem? Please describe.
Indexing speed of HNSW is not linear, so it might take a lot of time to build segment of size 2N and relatively small time o build 2 segments of size N + N.

Describe the solution you'd like
Limit maximal size of the segment to achieve faster indexing speed.
Merge optimizer should not perform merge with segments of size >= max_size_limit.

(Alternatively, it could check the size of resulting segment, but it might be a little harder to implement)

Describe alternatives you've considered
Check other existing OS implementations and investigate how do they solve the problem.

generall added a commit that referenced this issue Jan 18, 2022
* add parameters to optimizer config

* benchmark search speed in different segment sizes

* use constructor for FilteredScorer

* * Implement benchmarks for HNSW index search with different number of
  stored points
* Fix minor issue in HNSW graph edge assignment
* Update profiler with call-graph report generation
* Add profiling guide
* Add HNSW graph statistics test function (debug inly)

* limit resulting segment size in merge optimizer

* fix clippy

* stop the music

* fix clippy once again

* fmt once again
@generall
Copy link
Member Author

generall commented Feb 3, 2022

@generall generall closed this as completed Feb 3, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant