perf: parallelize CPU-heavy parquet metadata parsing in list_files_for_scan#21692
Closed
Dandandan wants to merge 2 commits intoapache:mainfrom
Closed
perf: parallelize CPU-heavy parquet metadata parsing in list_files_for_scan#21692Dandandan wants to merge 2 commits intoapache:mainfrom
list_files_for_scan#21692Dandandan wants to merge 2 commits intoapache:mainfrom
Conversation
…or_scan` Wraps the metadata fetch + statistics/ordering extraction inside `ParquetFormat::infer_stats_and_ordering` in `SpawnedTask::spawn_blocking` with `handle.block_on` so each call runs on a separate worker thread. `list_files_for_scan` already drives this via `.buffer_unordered(meta_fetch_concurrency)`, so concurrent per-file work now actually runs in parallel across threads instead of serializing the CPU-heavy parquet metadata decode/merge onto a single async task. Follows the same pattern as apache#19969 (parallelizing `infer_schema`). Part of apache#19971. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Which issue does this PR close?
list_files_for_scan#19971.Rationale for this change
On cold runs,
list_files_for_scanbottlenecks on a single thread — the heavy part is not file listing IO but CPU work inside parquet metadata decode/merge/statistics extraction. Profiling shows all of this collapsed onto one async task even thoughlist_files_for_scanalready drives per-file work with.buffer_unordered(meta_fetch_concurrency).buffer_unorderedpolls futures concurrently on a single task, so CPU-bound futures serialize.What changes are included in this PR?
Wrap the metadata fetch + statistics/ordering extraction in
ParquetFormat::infer_stats_and_orderingwithSpawnedTask::spawn_blocking(move || handle.block_on(...)), so each call runs on a separate worker thread. Combined with the existingbuffer_unordered(meta_fetch_concurrency)inlist_files_for_scan, we now get real parallelism across files.This follows the same pattern as #19969 (parallelizing
infer_schema).Trait signatures are unchanged; only the parquet implementation is touched.
Are these changes tested?
Covered by existing parquet / catalog-listing tests (
cargo test -p datafusion-datasource-parquet,cargo test -p datafusion-catalog-listing).Are there any user-facing changes?
No API changes. Cold-start listing of parquet tables with many files should be noticeably faster on multi-core systems.