You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@spark.apache.org by do...@apache.org on 2026/03/31 12:58:50 UTC
(spark) branch branch-4.1 updated: [SPARK-56279][CORE][4.1] Enable zero-copy sendfile for FileRegion in native Netty transports
This is an automated email from the ASF dual-hosted git repository.
dongjoon pushed a commit to branch branch-4.1
in repository https://gitbox.apache.org/repos/asf/spark.git
The following commit(s) were added to refs/heads/branch-4.1 by this push:
new c1d739c5a377 [SPARK-56279][CORE][4.1] Enable zero-copy sendfile for FileRegion in native Netty transports
c1d739c5a377 is described below
commit c1d739c5a3772c6121e9d2e427ac7a446a88ff9e
Author: yangjie01 <ya...@baidu.com>
AuthorDate: Tue Mar 31 05:58:30 2026 -0700
[SPARK-56279][CORE][4.1] Enable zero-copy sendfile for FileRegion in native Netty transports
### What changes were proposed in this pull request?
This PR modifies `MessageEncoder` to emit the header `ByteBuf` and `FileRegion` as **separate objects** in the outbound message list when the body is a `FileRegion` backed by `FileSegmentManagedBuffer`, instead of wrapping them together in a `MessageWithHeader`.
Previously, all messages with a body were unconditionally wrapped in `MessageWithHeader`. This caused native transports (EPOLL, KQUEUE) to fall into a generic `FileRegion.transferTo()` fallback path that copies data through user-space, bypassing the optimized `sendfile()` / `splice()` zero-copy path that Netty's native transports provide.
The split is only applied when the `ManagedBuffer` is a `FileSegmentManagedBuffer`, whose `release()` is a no-op, making it safe to emit the `FileRegion` independently of write lifecycle management. Other `ManagedBuffer` types (e.g., `BlockManagerManagedBuffer`) still use the `MessageWithHeader` wrapper because they perform resource cleanup in `release()` that must be tied to `MessageWithHeader.deallocate()`.
### Why are the changes needed?
When using native transports (AUTO/EPOLL on Linux), file-backed shuffle fetch performance was severely degraded compared to NIO mode. The root cause lies in how Netty's native transports dispatch `FileRegion` writes.
In `AbstractEpollStreamChannel.doWriteSingle()` (and the analogous KQueue path), Netty uses an `instanceof` check to choose between two write strategies:
https://github.com/netty/netty/blob/eeb5674526f0b49a142580686a5a9a7147ddadec/transport-classes-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java#L474-L493
```java
} else if (msg instanceof DefaultFileRegion) {
return writeDefaultFileRegion(in, (DefaultFileRegion) msg); // → socket.sendFile() (zero-copy)
} else if (msg instanceof FileRegion) {
return writeFileRegion(in, (FileRegion) msg); // → region.transferTo() (user-space copy)
}
```
- **`writeDefaultFileRegion()`** calls `socket.sendFile()`, which maps directly to the Linux `sendfile()` syscall — a true zero-copy path where data is transferred from the file page cache to the socket buffer entirely within the kernel, with no user-space copy.
https://github.com/netty/netty/blob/eeb5674526f0b49a142580686a5a9a7147ddadec/transport-classes-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java#L367-L386
```java
private int writeDefaultFileRegion(ChannelOutboundBuffer in, DefaultFileRegion region) throws Exception {
final long offset = region.transferred();
final long regionCount = region.count();
if (offset >= regionCount) {
in.remove();
return 0;
}
final long flushedAmount = socket.sendFile(region, region.position(), offset, regionCount - offset);
if (flushedAmount > 0) {
in.progress(flushedAmount);
if (region.transferred() >= regionCount) {
in.remove();
}
return 1;
} else if (flushedAmount == 0) {
validateFileRegion(region, offset);
}
return WRITE_STATUS_SNDBUF_FULL;
}
```
- **`writeFileRegion()`** falls back to `region.transferTo(WritableByteChannel)`, which writes data through a `SocketWritableByteChannel` wrapper — effectively a user-space copy path.
https://github.com/netty/netty/blob/eeb5674526f0b49a142580686a5a9a7147ddadec/transport-classes-epoll/src/main/java/io/netty/channel/epoll/AbstractEpollStreamChannel.java#L402-L420
```java
private int writeFileRegion(ChannelOutboundBuffer in, FileRegion region) throws Exception {
if (region.transferred() >= region.count()) {
in.remove();
return 0;
}
if (byteChannel == null) {
byteChannel = new EpollSocketWritableByteChannel();
}
final long flushedAmount = region.transferTo(byteChannel, region.transferred());
if (flushedAmount > 0) {
in.progress(flushedAmount);
if (region.transferred() >= region.count()) {
in.remove();
}
return 1;
}
return WRITE_STATUS_SNDBUF_FULL;
}
```
Spark's `MessageWithHeader extends AbstractFileRegion` (not `DefaultFileRegion`). When `MessageEncoder` wraps a `DefaultFileRegion` body inside `MessageWithHeader`, the resulting object is a generic `FileRegion` from Netty's perspective. This means Netty dispatches it to the `writeFileRegion()` fallback, which calls `MessageWithHeader.transferTo()`:
```java
// MessageWithHeader.java, line 121
if (body instanceof FileRegion fileRegion) {
writtenBody = fileRegion.transferTo(target, totalBytesTransferred - headerLength);
}
```
Here, even though the inner body is a `DefaultFileRegion`, its `transferTo()` is invoked with a `WritableByteChannel` (not a file descriptor), so the data is read from the file into a user-space buffer and then written to the socket — **the zero-copy opportunity is lost**.
By emitting the `DefaultFileRegion` directly into Netty's outbound buffer (instead of wrapping it in `MessageWithHeader`), Netty's native transport recognizes it via `instanceof DefaultFileRegion` and routes it to `socket.sendFile()`, restoring the zero-copy `sendfile()` path.
**Benchmark results (File-Backed Shuffle Fetch) show dramatic improvement:**
| Scenario | Before (ms) | After (ms) | Improvement |
|---|---|---|---|
| AUTO, sequential fetch (JDK17) | 2052 | 366 | **~5.6x faster** |
| AUTO, parallel fetch (JDK17) | 647 | 194 | **~3.3x faster** |
| AUTO, sequential fetch (JDK21) | 1304 | 347 | **~3.7x faster** |
| AUTO, parallel fetch (JDK21) | 489 | 186 | **~2.6x faster** |
With this fix, AUTO (EPOLL) mode now matches or exceeds NIO performance for file-backed shuffle fetches, achieving the expected ~1.0X relative performance (previously ~0.2X).
### Does this PR introduce _any_ user-facing change?
No. This is an internal optimization to the Netty transport layer. Users benefit from improved shuffle fetch performance when using native transports (the default on Linux) without any configuration changes.
### How was this patch tested?
- Updated `MergedBlockMetaSuccessSuite` to validate the new encoder output format (2 separate objects: `ByteBuf` header + `DefaultFileRegion`, instead of a single `MessageWithHeader`).
- Re-ran `NettyTransportBenchmark` across JDK 17, JDK 21 to confirm the performance improvement. Updated benchmark result files accordingly.
### Was this patch authored or co-authored using generative AI tooling?
Generated-by: Claude Code 4.6
Closes #55107 from LuciferYang/SPARK-56279-4.1.
Lead-authored-by: yangjie01 <ya...@baidu.com>
Co-authored-by: LuciferYang <Lu...@users.noreply.github.com>
Signed-off-by: Dongjoon Hyun <do...@apache.org>
---
.../spark/network/protocol/MessageEncoder.java | 23 +++++++--
.../protocol/MergedBlockMetaSuccessSuite.java | 24 +++++++---
.../NettyTransportBenchmark-jdk21-results.txt | 54 +++++++++++-----------
.../benchmarks/NettyTransportBenchmark-results.txt | 54 +++++++++++-----------
4 files changed, 92 insertions(+), 63 deletions(-)
diff --git a/common/network-common/src/main/java/org/apache/spark/network/protocol/MessageEncoder.java b/common/network-common/src/main/java/org/apache/spark/network/protocol/MessageEncoder.java
index 3f23c17939e6..c855d0eb6f52 100644
--- a/common/network-common/src/main/java/org/apache/spark/network/protocol/MessageEncoder.java
+++ b/common/network-common/src/main/java/org/apache/spark/network/protocol/MessageEncoder.java
@@ -22,12 +22,14 @@ import java.util.List;
import io.netty.buffer.ByteBuf;
import io.netty.channel.ChannelHandler;
import io.netty.channel.ChannelHandlerContext;
+import io.netty.channel.FileRegion;
import io.netty.handler.codec.MessageToMessageEncoder;
import org.apache.spark.internal.LogKeys;
import org.apache.spark.internal.SparkLogger;
import org.apache.spark.internal.SparkLoggerFactory;
import org.apache.spark.internal.MDC;
+import org.apache.spark.network.buffer.FileSegmentManagedBuffer;
/**
* Encoder used by the server side to encode server-to-client responses.
@@ -89,9 +91,24 @@ public final class MessageEncoder extends MessageToMessageEncoder<Message> {
assert header.writableBytes() == 0;
if (body != null) {
- // We transfer ownership of the reference on in.body() to MessageWithHeader.
- // This reference will be freed when MessageWithHeader.deallocate() is called.
- out.add(new MessageWithHeader(in.body(), header, body, bodyLength));
+ if (body instanceof FileRegion && in.body() instanceof FileSegmentManagedBuffer) {
+ // Emit header and FileRegion as separate objects so that native transports
+ // (EPOLL, KQUEUE) can apply zero-copy sendfile/splice on the FileRegion directly.
+ // When wrapped in MessageWithHeader, native transports fall into a generic
+ // FileRegion.transferTo() fallback that copies data through user-space, bypassing
+ // the optimized sendfile() path.
+ //
+ // This split is only safe when the ManagedBuffer is FileSegmentManagedBuffer,
+ // whose release() is a no-op. Other ManagedBuffer types (e.g.,
+ // BlockManagerManagedBuffer) perform resource cleanup in release() that must be
+ // tied to the write lifecycle via MessageWithHeader.deallocate().
+ out.add(header);
+ out.add(body);
+ } else {
+ // We transfer ownership of the reference on in.body() to MessageWithHeader.
+ // This reference will be freed when MessageWithHeader.deallocate() is called.
+ out.add(new MessageWithHeader(in.body(), header, body, bodyLength));
+ }
} else {
out.add(header);
}
diff --git a/common/network-common/src/test/java/org/apache/spark/network/protocol/MergedBlockMetaSuccessSuite.java b/common/network-common/src/test/java/org/apache/spark/network/protocol/MergedBlockMetaSuccessSuite.java
index 41b84f389587..36ad9851930d 100644
--- a/common/network-common/src/test/java/org/apache/spark/network/protocol/MergedBlockMetaSuccessSuite.java
+++ b/common/network-common/src/test/java/org/apache/spark/network/protocol/MergedBlockMetaSuccessSuite.java
@@ -28,6 +28,7 @@ import io.netty.buffer.ByteBuf;
import io.netty.buffer.ByteBufAllocator;
import io.netty.buffer.Unpooled;
import io.netty.channel.ChannelHandlerContext;
+import io.netty.channel.DefaultFileRegion;
import org.junit.jupiter.api.Assertions;
import org.junit.jupiter.api.Test;
import org.roaringbitmap.RoaringBitmap;
@@ -70,14 +71,25 @@ public class MergedBlockMetaSuccessSuite {
when(context.alloc()).thenReturn(ByteBufAllocator.DEFAULT);
MessageEncoder.INSTANCE.encode(context, expectedMeta, out);
- Assertions.assertEquals(1, out.size());
- MessageWithHeader msgWithHeader = (MessageWithHeader) out.remove(0);
+ // FileSegmentManagedBuffer.convertToNetty() returns a DefaultFileRegion,
+ // so MessageEncoder splits the output into header ByteBuf + FileRegion.
+ Assertions.assertEquals(2, out.size());
+ Assertions.assertInstanceOf(ByteBuf.class, out.get(0));
+ Assertions.assertInstanceOf(DefaultFileRegion.class, out.get(1));
+ ByteBuf headerBuf = (ByteBuf) out.remove(0);
+ DefaultFileRegion fileRegion = (DefaultFileRegion) out.remove(0);
- ByteArrayWritableChannel writableChannel =
- new ByteArrayWritableChannel((int) msgWithHeader.count());
- while (msgWithHeader.transfered() < msgWithHeader.count()) {
- msgWithHeader.transferTo(writableChannel, msgWithHeader.transfered());
+ int totalLen = headerBuf.readableBytes() + (int) fileRegion.count();
+ ByteArrayWritableChannel writableChannel = new ByteArrayWritableChannel(totalLen);
+ // Write header
+ writableChannel.write(headerBuf.nioBuffer());
+ // Write file region body
+ fileRegion.open();
+ while (fileRegion.transferred() < fileRegion.count()) {
+ fileRegion.transferTo(writableChannel, fileRegion.transferred());
}
+ fileRegion.release();
+ headerBuf.release();
ByteBuf messageBuf = Unpooled.wrappedBuffer(writableChannel.getData());
messageBuf.readLong(); // frame length
MessageDecoder.INSTANCE.decode(mock(ChannelHandlerContext.class), messageBuf, out);
diff --git a/core/benchmarks/NettyTransportBenchmark-jdk21-results.txt b/core/benchmarks/NettyTransportBenchmark-jdk21-results.txt
index 8722329cf447..d9077969ab7e 100644
--- a/core/benchmarks/NettyTransportBenchmark-jdk21-results.txt
+++ b/core/benchmarks/NettyTransportBenchmark-jdk21-results.txt
@@ -6,7 +6,7 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
RPC Latency (1 KB): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 KB payload 312 324 14 0.0 62497.5 1.0X
+1 KB payload 331 335 4 0.0 66108.2 1.0X
================================================================================================
@@ -17,7 +17,7 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
RPC Latency (64 KB): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-64 KB payload 435 442 8 0.0 86927.5 1.0X
+64 KB payload 432 445 10 0.0 86379.4 1.0X
================================================================================================
@@ -28,7 +28,7 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
RPC Latency (1 MB): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 MB payload 465 478 8 0.0 465416.7 1.0X
+1 MB payload 455 467 8 0.0 454621.2 1.0X
================================================================================================
@@ -39,7 +39,7 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
RPC Latency (16 MB): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-16 MB payload 800 832 29 0.0 7998535.2 1.0X
+16 MB payload 798 802 3 0.0 7982304.4 1.0X
================================================================================================
@@ -50,10 +50,10 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
Concurrent RPC Throughput: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 client(s) 1264 1270 8 0.0 63217.5 1.0X
-4 client(s) 431 438 9 0.0 21569.1 2.9X
-8 client(s) 300 307 11 0.1 14983.0 4.2X
-16 client(s) 267 274 9 0.1 13371.1 4.7X
+1 client(s) 1246 1287 37 0.0 62282.4 1.0X
+4 client(s) 404 419 15 0.0 20185.8 3.1X
+8 client(s) 269 297 25 0.1 13440.8 4.6X
+16 client(s) 259 268 8 0.1 12953.1 4.8X
================================================================================================
@@ -64,8 +64,8 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
IOMode Comparison: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-NIO (8 clients) 553 658 151 0.0 27631.8 1.0X
-AUTO (8 clients) 490 502 16 0.0 24503.3 1.1X
+NIO (8 clients) 474 485 16 0.0 23677.9 1.0X
+AUTO (8 clients) 460 485 28 0.0 22991.0 1.0X
================================================================================================
@@ -76,11 +76,11 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
Server Thread Scaling: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-2 server threads 266 275 8 0.1 13285.6 1.0X
-4 server threads 236 242 5 0.1 11804.1 1.1X
-8 server threads 238 273 31 0.1 11885.5 1.1X
-16 server threads 261 276 24 0.1 13028.2 1.0X
-32 server threads 260 276 16 0.1 13003.7 1.0X
+2 server threads 276 283 7 0.1 13814.3 1.0X
+4 server threads 221 244 22 0.1 11029.7 1.3X
+8 server threads 240 260 23 0.1 11986.8 1.2X
+16 server threads 248 256 9 0.1 12379.9 1.1X
+32 server threads 257 270 15 0.1 12845.1 1.1X
================================================================================================
@@ -91,9 +91,9 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
Multi-Connection Throughput: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 conn(s), 4 threads 1730 1754 23 0.0 346078.1 1.0X
-2 conn(s), 4 threads 1192 1234 41 0.0 238362.5 1.5X
-4 conn(s), 4 threads 1230 1250 19 0.0 246007.4 1.4X
+1 conn(s), 4 threads 1688 1773 127 0.0 337503.3 1.0X
+2 conn(s), 4 threads 1210 1274 96 0.0 241972.1 1.4X
+4 conn(s), 4 threads 1140 1147 6 0.0 227969.5 1.5X
================================================================================================
@@ -104,9 +104,9 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
Async Write Throughput: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 KB async burst 45 46 1 0.1 9009.2 1.0X
-64 KB async burst 144 209 109 0.0 28863.1 0.3X
-1 MB async burst 1675 1721 80 0.0 334900.9 0.0X
+1 KB async burst 46 49 2 0.1 9214.1 1.0X
+64 KB async burst 130 138 12 0.0 25985.5 0.4X
+1 MB async burst 1598 1602 3 0.0 319655.6 0.0X
================================================================================================
@@ -117,8 +117,8 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
16 MB Block Transfer: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-Sequential sends 788 813 23 0.0 7875409.8 1.0X
-4-thread parallel sends 448 457 8 0.0 4480383.9 1.8X
+Sequential sends 798 813 13 0.0 7981232.3 1.0X
+4-thread parallel sends 473 485 10 0.0 4734290.2 1.7X
================================================================================================
@@ -129,9 +129,9 @@ OpenJDK 64-Bit Server VM 21.0.10+7-LTS on Linux 6.17.0-1008-azure
Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
File-Backed Shuffle Fetch: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-NIO, sequential fetch 346 356 14 0.0 3455769.0 1.0X
-NIO, parallel fetch (4 clients) 209 217 12 0.0 2085147.0 1.7X
-AUTO, sequential fetch 1304 1326 28 0.0 13042655.6 0.3X
-AUTO, parallel fetch (4 clients) 489 600 97 0.0 4888981.6 0.7X
+NIO, sequential fetch 353 356 3 0.0 3525033.6 1.0X
+NIO, parallel fetch (4 clients) 193 231 61 0.0 1931644.9 1.8X
+AUTO, sequential fetch 347 350 3 0.0 3467489.5 1.0X
+AUTO, parallel fetch (4 clients) 186 196 8 0.0 1860609.7 1.9X
diff --git a/core/benchmarks/NettyTransportBenchmark-results.txt b/core/benchmarks/NettyTransportBenchmark-results.txt
index d70fc947936e..9f19fced6cad 100644
--- a/core/benchmarks/NettyTransportBenchmark-results.txt
+++ b/core/benchmarks/NettyTransportBenchmark-results.txt
@@ -6,7 +6,7 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
RPC Latency (1 KB): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 KB payload 507 520 13 0.0 101304.2 1.0X
+1 KB payload 508 533 17 0.0 101559.9 1.0X
================================================================================================
@@ -17,7 +17,7 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
RPC Latency (64 KB): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-64 KB payload 699 721 21 0.0 139776.4 1.0X
+64 KB payload 729 780 30 0.0 145800.9 1.0X
================================================================================================
@@ -28,7 +28,7 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
RPC Latency (1 MB): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 MB payload 862 926 59 0.0 862232.2 1.0X
+1 MB payload 837 896 61 0.0 836593.9 1.0X
================================================================================================
@@ -39,7 +39,7 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
RPC Latency (16 MB): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-16 MB payload 1267 1313 54 0.0 12670789.5 1.0X
+16 MB payload 1249 1323 52 0.0 12486866.6 1.0X
================================================================================================
@@ -50,10 +50,10 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
Concurrent RPC Throughput: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 client(s) 1951 1976 21 0.0 97572.6 1.0X
-4 client(s) 663 724 53 0.0 33135.6 2.9X
-8 client(s) 489 511 22 0.0 24432.6 4.0X
-16 client(s) 453 456 3 0.0 22661.9 4.3X
+1 client(s) 2091 2115 21 0.0 104534.5 1.0X
+4 client(s) 760 790 26 0.0 37996.5 2.8X
+8 client(s) 557 566 10 0.0 27855.0 3.8X
+16 client(s) 461 478 26 0.0 23068.9 4.5X
================================================================================================
@@ -64,8 +64,8 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
IOMode Comparison: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-NIO (8 clients) 756 800 59 0.0 37803.0 1.0X
-AUTO (8 clients) 879 894 13 0.0 43947.8 0.9X
+NIO (8 clients) 785 864 88 0.0 39253.3 1.0X
+AUTO (8 clients) 861 868 6 0.0 43070.4 0.9X
================================================================================================
@@ -76,11 +76,11 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
Server Thread Scaling: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-2 server threads 459 472 12 0.0 22963.4 1.0X
-4 server threads 392 430 54 0.1 19612.1 1.2X
-8 server threads 427 438 17 0.0 21330.4 1.1X
-16 server threads 445 456 18 0.0 22247.3 1.0X
-32 server threads 457 508 46 0.0 22844.3 1.0X
+2 server threads 479 503 24 0.0 23938.2 1.0X
+4 server threads 413 429 25 0.0 20659.7 1.2X
+8 server threads 455 468 14 0.0 22765.9 1.1X
+16 server threads 455 486 34 0.0 22740.7 1.1X
+32 server threads 473 513 34 0.0 23672.8 1.0X
================================================================================================
@@ -91,9 +91,9 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
Multi-Connection Throughput: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 conn(s), 4 threads 1688 1791 155 0.0 337699.8 1.0X
-2 conn(s), 4 threads 1242 1546 272 0.0 248330.3 1.4X
-4 conn(s), 4 threads 1625 1687 62 0.0 324991.8 1.0X
+1 conn(s), 4 threads 1773 1841 62 0.0 354542.2 1.0X
+2 conn(s), 4 threads 1365 1554 167 0.0 272904.6 1.3X
+4 conn(s), 4 threads 1655 1694 37 0.0 330985.4 1.1X
================================================================================================
@@ -104,9 +104,9 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
Async Write Throughput: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-1 KB async burst 37 49 18 0.1 7305.3 1.0X
-64 KB async burst 141 149 7 0.0 28249.7 0.3X
-1 MB async burst 1319 1509 165 0.0 263898.9 0.0X
+1 KB async burst 73 76 3 0.1 14634.8 1.0X
+64 KB async burst 150 159 9 0.0 29915.2 0.5X
+1 MB async burst 1337 1462 173 0.0 267363.9 0.1X
================================================================================================
@@ -117,8 +117,8 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
16 MB Block Transfer: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-Sequential sends 632 825 322 0.0 6319420.9 1.0X
-4-thread parallel sends 446 451 4 0.0 4461770.8 1.4X
+Sequential sends 655 1135 416 0.0 6548597.0 1.0X
+4-thread parallel sends 452 464 11 0.0 4524357.9 1.4X
================================================================================================
@@ -129,9 +129,9 @@ OpenJDK 64-Bit Server VM 17.0.18+8-LTS on Linux 6.17.0-1008-azure
AMD EPYC 7763 64-Core Processor
File-Backed Shuffle Fetch: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-NIO, sequential fetch 367 376 11 0.0 3665068.0 1.0X
-NIO, parallel fetch (4 clients) 199 200 1 0.0 1992404.2 1.8X
-AUTO, sequential fetch 2052 2118 94 0.0 20522297.2 0.2X
-AUTO, parallel fetch (4 clients) 647 658 13 0.0 6466373.1 0.6X
+NIO, sequential fetch 373 381 7 0.0 3734242.2 1.0X
+NIO, parallel fetch (4 clients) 202 210 11 0.0 2024154.0 1.8X
+AUTO, sequential fetch 366 375 9 0.0 3660147.3 1.0X
+AUTO, parallel fetch (4 clients) 194 201 9 0.0 1944456.7 1.9X
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@spark.apache.org
For additional commands, e-mail: commits-help@spark.apache.org