perf(store): batch reads #331
Merged
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
Wondertan
commented
Aug 1, 2025
Comment on lines
+766
to
+769
| bds, ok := s.ds.Children()[0].(datastore.Batching) | ||
| if !ok { | ||
| return ctx, func() error { return nil } | ||
| } |
There was a problem hiding this comment.
A note for an advanced reader: we need to unwrap the underlying datastore from the keytransformer one, otherwise key transformation is going to happen twice in the datastore itself and the tx wrapper.
Basically, a combination of keytransformer and contextds may lead to unexpected and silent double transformation, which is annoying to debug. There might be a general way to untangle these on the library side, but this is the simplest solution on our side I could think of.
walldiss
reviewed
Aug 1, 2025
walldiss
approved these changes
Aug 1, 2025
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Outdated suggestions cannot be applied.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.
This PR shaves another 30s when we prune the whole chain. It achieves this by applying the same technique as in #328, but for reads. Previously, the underlying datastore would create a read-only transaction for every singular read. As shown in pyroscope profiles, this puts pressure on badger as creating new transactions ain't cheap and involves complex synchronization to ensure atomicity between parallel transactions.
With this insight in mind, I also integrated read batching in other places where this must be beneficial: advancing, receding, range, and singular header reads.
More noticeable changes: