Repository navigation
bench: measure large string IN lists on the DataFusion Iceberg scan - #130
Draft
alexanderbianchi wants to merge 2 commits into
Draft
alexanderbianchi wants to merge 2 commits into
alexanderbianchi wants to merge 2 commits into
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Benchmark
This Criterion benchmark measures how long DataFusion takes to plan and collect a query of the form
SELECT resource_key FROM taxi WHERE resource_key IN (...). It builds a seven-file Iceberg taxi table once, generates a distinct 63–68 byte string key for each trip, and checks the exact matching keys before timing each list size. The companion Iceberg reader change starts using a hash set at 32 distinct string literals. apache/iceberg-rust#3739Results
At 400 keys, mean query time fell from 155.08 ms to 14.91 ms (10.40× faster). Both readers returned the same 400 keys out of 175,000 trips.
These are Criterion’s mean wall-clock estimates from 20 samples per case on an Apple M5 Pro, using the same seven January 2024 NYC taxi source files for both builds. Each run used one local query at a time with four Tokio workers and four DataFusion target partitions. The 95% confidence intervals at 400 keys were 149.88–163.15 ms before and 14.66–15.23 ms after. Taxi fixture
Run it
The default uses the repository’s deterministic taxi fixture. To reproduce the table above, set
TAXI_DATA_DIRto the archived fixture’sdata/directory with its Git LFS files materialized. Both runs used DataFusion 55.2.0, Arrow 59.3.0 and reader baseline9354b42. The second build added only the hash-set commit. Baseline reader · Reader changeWith the baseline reader selected, save a Criterion baseline. After selecting the patched reader, compare in the same target directory.