vanzin commented on a change in pull request #23602:
[SPARK-26674][CORE]Consolidate CompositeByteBuf when reading large frame
URL: https://github.com/apache/spark/pull/23602#discussion_r255275374
##########
File path:
common/network-common/src/test/java/org/apache/spark/network/util/TransportFrameDecoderSuite.java
##########
@@ -47,6 +47,76 @@ public void testFrameDecoding() throws Exception {
verifyAndCloseDecoder(decoder, ctx, data);
}
+ @Test
+ public void testConsolidationForDecodingNonFullyWrittenByteBuf() {
+ TransportFrameDecoder decoder = new TransportFrameDecoder();
+ ChannelHandlerContext ctx = mock(ChannelHandlerContext.class);
+ List<ByteBuf> retained = new ArrayList<>();
+ when(ctx.fireChannelRead(any())).thenAnswer(in -> {
+ ByteBuf buf = (ByteBuf) in.getArguments()[0];
+ retained.add(buf);
+ return null;
+ });
+ ByteBuf data1 = Unpooled.buffer(1024 * 1024);
+ data1.writeLong(1024 * 1024 + 8);
+ data1.writeByte(127);
+ ByteBuf data2 = Unpooled.buffer(1024 * 1024);
+ for (int i = 0; i < 1024 * 1024 - 1; i++) {
+ data2.writeByte(128);
+ }
+ int orignalCapacity = data1.capacity() + data2.capacity();
+ try {
+ decoder.channelRead(ctx, data1);
+ decoder.channelRead(ctx, data2);
+ assertEquals(1, retained.size());
+ assert(retained.get(0).capacity() < orignalCapacity);
+ } catch (Exception e) {
+ release(data1);
+ release(data2);
+ }
+ }
+
+ @Test
+ public void testConsolidationPerf() throws Exception {
+ long[] testingConsolidateThresholds = new long[] {
+ 1024 * 1024, 5 * 1024 * 1024, 10 * 1024 * 1024, 20 * 1024 * 1024,
+ 30 * 1024 * 1024, 50 * 1024 * 1024, 80 * 1024 * 1024, 100 * 1024 *
1024,
+ 300 * 1024 * 1024, 500 * 1024 * 1024, Long.MAX_VALUE };
+ for (long threshold : testingConsolidateThresholds) {
+ TransportFrameDecoder decoder = new TransportFrameDecoder(threshold);
+ ChannelHandlerContext ctx = mock(ChannelHandlerContext.class);
+ List<ByteBuf> retained = new ArrayList<>();
+ when(ctx.fireChannelRead(any())).thenAnswer(in -> {
+ ByteBuf buf = (ByteBuf) in.getArguments()[0];
+ retained.add(buf);
+ return null;
+ });
+
+ try {
+ long start = System.currentTimeMillis();
+ ByteBuf buf = Unpooled.buffer(8);
+ buf.writeLong(8 + 1024 * 1024 * 1000);
+ decoder.channelRead(ctx, buf);
+ for (int i = 0; i < 1000; i++) {
Review comment:
Is the goal here to have a test or a realistic benchmark?
If it's a test, I'd have more randomized write sizes. Have a target number
of bytes and keep on writing until you reach that. That's better to try to
exercise corner cases.
If it's a benchmark, I'd probably choose a smaller write buffer; either 32k
or 64k. That's probably going to be closer to what you see when reading a large
shuffle over the network.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
