bench: io_uring-backed LocalFileSystem for dfbench (Linux)#21793
Closed
Dandandan wants to merge 2 commits intoapache:mainfrom
Closed
bench: io_uring-backed LocalFileSystem for dfbench (Linux)#21793Dandandan wants to merge 2 commits intoapache:mainfrom
Dandandan wants to merge 2 commits intoapache:mainfrom
Conversation
Adds `UringLocalFileSystem`, an `ObjectStore` that routes byte-range reads through a dedicated io_uring driver thread and delegates all other operations to an inner `LocalFileSystem`. Wired into `CommonOpt::build_runtime` and registered for `file:///` by default on Linux; opt out with `DATAFUSION_IO_URING=0`. Non-Linux targets are unaffected — the module and the `io-uring` dep are both gated on `target_os = "linux"`. Design: - `UringSubmitter` owns an `mpsc::UnboundedSender<Cmd>`; any tokio task calls `submit_read` (sync) to enqueue an `IORING_OP_READ`, getting back a `oneshot::Receiver<Result<Bytes>>`. - A dedicated `io-uring-driver` thread owns the `IoUring` instance. Each iteration drains the mpsc up to the submission queue's free slots, `submit_and_wait(1)` to flush SQEs and block for at least one CQ entry when work is outstanding, then drains CQ entries and fires the oneshots. When nothing is in flight it `blocking_recv()` for the next command instead of spinning. - `get_ranges` submits all N ranges into the mpsc synchronously before awaiting any of them, so the driver sees the whole batch in one drain and hands the kernel all N at once — giving the device queue depth > 1 for cold reads instead of the one-pread-at-a-time pattern that `LocalFileSystem::get_ranges` produces today. - Buffers (`Box<[u8]>`) and the `Arc<File>` keeping the fd open are retained in the driver's `in_flight` map until the corresponding CQ entry arrives, so the kernel never writes into freed memory or a closed fd. Rough edges (called out in module-level docs): - No fd cache (one `open(2)` per `get_ranges` call, same as today). - No registered buffers / `IORING_OP_READV` — one SQE per range, heap allocation per op. - No `IORING_OP_ASYNC_CANCEL` on dropped-future cancellation; the submission runs to completion and its result is discarded. - Metrics / tracing not yet plumbed in. Intended as a first draft for benchmarking; good enough to A/B against the stock `LocalFileSystem` on ClickBench-style workloads. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Contributor
Author
|
run benchmarks |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing io-uring-local-fs (8844eb9) to 9a1ed57 (merge-base) diff using: clickbench_partitioned File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing io-uring-local-fs (8844eb9) to 9a1ed57 (merge-base) diff using: tpcds File an issue against this benchmark runner |
|
Benchmark for this request failed. Last 20 lines of output: Click to expandFile an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing io-uring-local-fs (8844eb9) to 9a1ed57 (merge-base) diff using: tpch File an issue against this benchmark runner |
|
Benchmark for this request failed. Last 20 lines of output: Click to expandFile an issue against this benchmark runner |
1 similar comment
|
Benchmark for this request failed. Last 20 lines of output: Click to expandFile an issue against this benchmark runner |
The Debug derive on UringLocalFileSystem transitively requires Debug on its fields; UringSubmitter is one of them, and tokio's mpsc::UnboundedSender<T> already impls Debug regardless of T, so a plain derive is enough. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Contributor
Author
|
run benchmarks |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing io-uring-local-fs (fbeeee2) to 9a1ed57 (merge-base) diff using: tpch File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing io-uring-local-fs (fbeeee2) to 9a1ed57 (merge-base) diff using: clickbench_partitioned File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing io-uring-local-fs (fbeeee2) to 9a1ed57 (merge-base) diff using: tpcds 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 |
|
Benchmark for this request hit the 7200s job deadline before finishing. Benchmarks requested: Kubernetes messageFile 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
When profiling DataFusion's local parquet reads under ClickBench,
object_store::LocalFileSystem::get_rangesserializes all range reads inside a singlespawn_blockingtask:One blocking thread, N sequential
seek + read_exactpairs. On NVMe devices with meaningful queue-depth capability, and on cold-cache reads, this leaves a lot of parallelism unused — the kernel block layer would happily service many concurrent reads if we asked it to.This PR adds a benchmark-only alternative
ObjectStore(no changes todatafusion/orobject_store) that routes the range reads through anio_uringsubmission queue, so N preads become N concurrent kernel-side operations. It's intended as a tool for A/B measurement rather than a production-quality replacement.What changes are included in this PR?
benchmarks/src/util/uring_local_fs.rs(new, ~480 lines): aUringLocalFileSystemimplementingobject_store::ObjectStore. It owns aninner: LocalFileSystemfor non-read ops and a dedicatedio-uring-driverOS thread that owns theIoUringinstance and the submission/completion loop.benchmarks/src/util/mod.rs: registers the module under#[cfg(target_os = \"linux\")].benchmarks/src/util/options.rs:CommonOpt::build_runtimeregistersUringLocalFileSystemforfile:///by default on Linux, withDATAFUSION_IO_URING=0as the opt-out. Layers with--simulate-latencyas expected (LatencyObjectStorewraps the uring store).benchmarks/Cargo.toml:io-uring = \"0.7\"added under[target.'cfg(target_os = \"linux\")'.dependencies], so non-Linux targets don't pull it in.Driver shape:
submit_read(Arc<File>, offset, len)— a sync fn — which sends aCmd::Readover an mpsc and returns aoneshot::Receiver<io::Result<Bytes>>. This is sync on purpose:get_rangesenqueues all N ranges before awaiting any of them, so the driver sees the whole batch in onetry_recvdrain.submit_and_wait(1)to flush and block for at least one completion when work is outstanding, then drains the CQ and fires the oneshots. Idles withblocking_recv()when empty.Box<[u8]>) and the keep-aliveArc<File>live in the driver'sin_flightmap until the corresponding CQ arrives — the kernel never writes into freed memory or a closed fd.Known rough edges (documented in the module header):
open(2)perget_rangescall (same as today).IORING_OP_READV— one SQE per range, heap allocation per op.IORING_OP_ASYNC_CANCELon dropped-future cancellation; the submission runs to completion and its result is discarded.Not included in this PR: any change to
object_storeordatafusioncore, or any production path. All non-Linux users get the stockLocalFileSystemvia the existing cfg-gated code.Are these changes tested?
cargo check -p datafusion-benchmarksandcargo clippy -p datafusion-benchmarks --all-targets -- -D warningspass on macOS (Linux module is cfg-d out)../target/release-nonlto/dfbench clickbench --iterations 3 --path <hits_partitioned> --queries-path benchmarks/queries/clickbench/querieswith and withoutDATAFUSION_IO_URING=0is the expected first validation.Are there any user-facing changes?
Only within
dfbenchon Linux:Using io_uring-backed LocalFileSystemso it's visible which backend is in effect.DATAFUSION_IO_URING=0in the environment restores the stockLocalFileSystem.No API changes. No changes to any crate that downstream users depend on.