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

[ENH] Evict other versions of hnsw index of the collection when another version is fetched #2707

Conversation

sanketkedia
Copy link
Contributor

@sanketkedia sanketkedia commented Aug 22, 2024

Description of changes

Summarize the changes made by this PR.
It was seen that query nodes OOM in load tests. One area where we could optimize the memory footprint is the HNSW index cache. This PR restricts the cache to keep one entry per collection. Ideally, we would like the index version of this entry to be of the latest but there is no guarantee with this PR. Once the versioning changes land, we could improve this. For e.g. one bad case could be:
For the same collection:

  1. get index version v1
  2. get index version v2 (> v1)
  3. get index version v1 (can happen due to an inflight query that started before compaction of v2 occured) -- this will evict v2 even though it is more recent and will be used again in future

Test plan

  • Tests pass locally with pytest for python, yarn test for js, cargo test for rust

Documentation Changes

None

Copy link

Reviewer Checklist

Please leverage this checklist to ensure your code review is thorough before approving

Testing, Bugs, Errors, Logs, Documentation

  • Can you think of any use case in which the code does not behave as intended? Have they been tested?
  • Can you think of any inputs or external events that could break the code? Is user input validated and safe? Have they been tested?
  • If appropriate, are there adequate property based tests?
  • If appropriate, are there adequate unit tests?
  • Should any logging, debugging, tracing information be added or removed?
  • Are error messages user-friendly?
  • Have all documentation changes needed been made?
  • Have all non-obvious changes been commented?

System Compatibility

  • Are there any potential impacts on other parts of the system or backward compatibility?
  • Does this change intersect with any items on our roadmap, and if so, is there a plan for fitting them together?

Quality

  • Is this code of a unexpectedly high quality (Readability, Modularity, Intuitiveness)

Copy link

Please tag your PR title with one of: [ENH | BUG | DOC | TST | BLD | PERF | TYP | CLN | CHORE]. See https://docs.trychroma.com/contributing#contributing-code-and-ideas

Copy link
Contributor Author

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @sanketkedia and the rest of your teammates on Graphite Graphite

@sanketkedia sanketkedia changed the title Evict other versions of hnsw index of the collection when another version is available [ENH] Evict other versions of hnsw index of the collection when another version is fetched Aug 22, 2024
@sanketkedia sanketkedia marked this pull request as ready for review August 22, 2024 17:45
Copy link
Collaborator

@HammadB HammadB left a comment

Choose a reason for hiding this comment

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

I think this is fine until we have version included, and we can be refcount aware / do a deferred eviction.

Is it possible to add some basic documentation in the comments describing this behavior and the motivation?

@sanketkedia sanketkedia merged commit 9661326 into main Aug 22, 2024
68 checks passed
spikechroma pushed a commit that referenced this pull request Aug 22, 2024
…er version is fetched (#2707)

## Description of changes

*Summarize the changes made by this PR.*
It was seen that query nodes OOM in load tests. One area where we could
optimize the memory footprint is the HNSW index cache. This PR restricts
the cache to keep one entry per collection. Ideally, we would like the
index version of this entry to be of the latest but there is no
guarantee with this PR. Once the versioning changes land, we could
improve this. For e.g. one bad case could be:
For the same collection:
1. get index version v1
2. get index version v2 (> v1)
3. get index version v1 (can happen due to an inflight query that
started before compaction of v2 occured) -- this will evict v2 even
though it is more recent and will be used again in future

## Test plan
- [x] Tests pass locally with `pytest` for python, `yarn test` for js,
`cargo test` for rust

## Documentation Changes
None
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

Successfully merging this pull request may close these issues.

3 participants