[LuceneOnFaiss] Apply monotonic increasing integer encoding to FAISS HNSW and IdMapIndex. #2609
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
…dex. Signed-off-by: Dooyong Kim <[email protected]>
0ctopus13prime
requested review from
heemin32,
navneet1v,
VijayanB,
vamshin,
jmazanec15,
naveentatikonda,
junqiu-lei,
martin-gaievski,
ryanbogan,
luyuncheng,
shatejas and
Vikasht34
as code owners
5 tasks
jmazanec15
reviewed
Mar 20, 2025
src/main/java/org/opensearch/knn/memoryoptsearch/faiss/MonotonicIntegerSequenceEncoder.java
Show resolved
Hide resolved
|
@shatejas @jmazanec15 Could you approve this PR if it looks good on you? :) |
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Outdated suggestions cannot be applied.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.
Description
This PR introduces a monotonic integer sequence encoding scheme for doc IDs in IdMapIndex and offsets in FAISS HNSW.
Previously, both were stored as long[], consuming O(N) memory where N is the total number of vectors. For example, with a 10M dataset, they occupied 228MB (76MB + 152MB). Using monotonic encoding, this can be reduced to a few hundred KB. Lucene already uses this encoding scheme Link
The key difference is that Lucene applies encoding during segment creation, whereas we do it at load time. However, encoding is extremely fast, on my local setup it takes about 3.2 seconds for 100M integers.
A warm-up API could further optimize this by preloading data before search, though this hasn't been raised yet, but it is part of future PR.
Related Issues
Resolves #[Issue number to be closed when this PR is merged]
RFC : #2401
Check List
--signoff.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.