CNDB-15527: Add config to use FusedPQ #2042
Open
Conversation
This file contains hidden or 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
Checklist before you submit for review
|
michaeljmarshall
force-pushed
the
revive-fused-adc
branch
from
e63f53d to
5e02971
Compare
michaeljmarshall
force-pushed
the
revive-fused-adc
branch
from
5e02971 to
7c3c804
Compare
|
|
||
| VectorCompression.CompressionType compressionType = VectorCompression.CompressionType.values()[reader.readByte()]; | ||
| if (features.contains(FeatureId.FUSED_ADC)) | ||
| if (features.contains(FeatureId.FUSED_PQ)) |
Member
Author
There was a problem hiding this comment.
@marianotepper - I just noticed that the features map already has logic that loads the ProductQuantization, meaning this branch currently keeps two identical maps in memory. I think it'd make sense to possibly expose the features map in the OnDiskGraphIndex so we can remove the duplicate cost. Any reason we can't do that?
Member
Author
There was a problem hiding this comment.
Looks like we actually already use the one from the header, I just didn't catch it. I think we'll be able to get rid of it with a little extra work in CC.
built from the branch in this PR: datastax/jvector#588
This is a hack because it increases memory utilization and may impact other threads in the CC host. However, this will help unblock certain test scenarios.
Improve hashCode of the Key to prevent collisions Implement Comparable<Key> in case the ConcurrentHashMap falls back in to Tree mode (cherry picked from commit 091b636)
|
I have cherry-pick this into this feature branch: #2156 |
|
michaeljmarshall
changed the title
❌ Build ds-cassandra-pr-gate/PR-2042 rejected by Butler2 regressions found Found 2 new test failures
Found 2 known test failures |
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.
What is the issue
https://github.com/riptano/cndb/issues/15527
What does this PR fix and why was it fixed
Adds ability to build vector indexes using Fused PQ. Enabled by default. Can be disabled via
cassandra.sai.vector.enable_fusedconfig param. When enabled, fused graphs store the product quantization in the graph on disk, removing the need to store it in memory, and unlocking new scale for vector indexes.