perf(repartition): batch reservation + sends via per-partition local buffers#21677
Closed
Dandandan wants to merge 1 commit intoapache:mainfrom
Closed
perf(repartition): batch reservation + sends via per-partition local buffers#21677Dandandan wants to merge 1 commit intoapache:mainfrom
Dandandan wants to merge 1 commit intoapache:mainfrom
Conversation
…buffers Accumulate batches into per-partition local buffers in each input task and flush only when a buffer reaches 256 KiB. This amortizes the memory reservation + channel overhead: each flush performs a single `try_grow` call and spill-check for the entire buffered size, instead of once per tiny sub-batch. When hash-partitioning an input batch into N output partitions each output partition only gets ~input_size / N bytes per input batch. With many output partitions these sub-batches are very small and hitting the reservation Mutex (plus the backpressure gate mutex) on every one is a significant contention point that reduces CPU utilization for parallel-partition workloads like ClickBench. Semantics are preserved: - Memory batches and spill fallback behavior are unchanged. - If the receiver drops mid-flush, the full reserved amount is released. - End-of-input triggers a flush of any remaining buffered batches. No user-visible behavior change besides batching granularity. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Contributor
Author
|
run benchmarks |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing repartition-local-buffers (f2270ad) to 3b5008a (merge-base) diff using: clickbench_partitioned File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing repartition-local-buffers (f2270ad) to 3b5008a (merge-base) diff using: tpcds File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing repartition-local-buffers (f2270ad) to 3b5008a (merge-base) diff using: tpch File an issue against this benchmark runner |
|
🤖 Benchmark completed (GKE) | trigger Instance: CPU Details (lscpu)Details
Resource Usagetpcds — base (merge-base)
tpcds — branch
File an issue against this benchmark runner |
|
🤖 Benchmark completed (GKE) | trigger Instance: CPU Details (lscpu)Details
Resource Usageclickbench_partitioned — base (merge-base)
clickbench_partitioned — branch
File an issue against this benchmark runner |
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
2 participants
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?
Rationale for this change
In
RepartitionExec::pull_from_input, each input task partitions an input batch and sends the resulting sub-batches to output channels. On every sub-batch, the input task:Mutexand callstry_growchannel.state.lock()and potentially the gate mutex)When hash-partitioning an input batch of size
SintoNoutput partitions, each output gets a sub-batch of size~S/N. With many output partitions (e.g. 16 cores → 16 output partitions), these sub-batches are very small, so the per-batch synchronization cost dominates.What changes are included in this PR?
Each input task now accumulates partitioned sub-batches in per-output-partition local buffers (
Vec<PartitionBuffer>) and only flushes when a buffer reachesFLUSH_THRESHOLD_BYTES(256 KiB). Each flush performs a singletry_growcall for the total buffered size instead of once per sub-batch. Iftry_growfails, all buffered batches for that partition are spilled together.End-of-input triggers a final flush of any remaining buffered data per partition.
Are these changes tested?
Covered by the existing repartition test suite (41 tests pass), including the spilling, delayed-stream, dropped-output-stream, and ordering-preservation tests.
Are there any user-facing changes?
No. The change only affects batching granularity inside
RepartitionExec; memory semantics, spilling behavior, and output ordering are preserved.🤖 Generated with Claude Code