GitHub user Beihao-Zhou added a comment to the discussion: Vector Search HNSW 
Indexing Encoding

Thanks for sharing the advice! HNSW's first draft PR is here: 
https://github.com/apache/kvrocks/pull/2368
The indexing encoding is defined under 
[search_encoding.h](https://github.com/apache/kvrocks/pull/2368/files#diff-0386c0363702f185ec28a51eed74321378c00c2ccee7ad64a8060bf3523eebc7R219)
 within the PR. 

The implementation of insertion/search follows pretty much as 
[bustub-vectordb](https://github.com/skyzh/bustub-vectordb), but basically a 
modified on-disk version of the original [HNSW Algorithm 
Paper](https://arxiv.org/pdf/1603.09320). 

For block cache, I see that it's been managed by the storage layer when RocksDB 
is open 
[here](https://github.com/apache/kvrocks/blob/7d48490ba1890d9bb0510c11350d7ae1a1f14dd0/src/storage/storage.cc#L287).
 I think for vector search, RocksDB might need to be carefully tuned. 

GitHub link: 
https://github.com/apache/kvrocks/discussions/2316#discussioncomment-9823337

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to