streams: 5% throughput gain when sending small chunks #4354

mcollina · 2015-12-19T17:11:33Z

Improves the performance when moving small buffers by 5%, and it adds a benchmark to avoid regression in that area. In all other cases it is equally performant current master. Some cases turns out to be faster with this patch, especially in cluster under high load.
It also tidy up the clearBuffer method, moving most of the not-necessary code out from the hot code path.

Full performance results available at: https://gist.github.com/mcollina/717c35ad07d15710b6b9.

These are the relevant lines, before:

net/net-c2s-cork.js
net/net-c2s-cork.js len=4 type=buf dur=5: 0.07900
net/net-c2s-cork.js len=8 type=buf dur=5: 0.16693
net/net-c2s-cork.js len=16 type=buf dur=5: 0.32318
net/net-c2s-cork.js len=32 type=buf dur=5: 0.56029
net/net-c2s-cork.js len=64 type=buf dur=5: 1.31117
net/net-c2s-cork.js len=128 type=buf dur=5: 2.19901
net/net-c2s-cork.js len=512 type=buf dur=5: 4.64193
net/net-c2s-cork.js len=1024 type=buf dur=5: 5.46635

and after:

net/net-c2s-cork.js
net/net-c2s-cork.js len=4 type=buf dur=5: 0.08748
net/net-c2s-cork.js len=8 type=buf dur=5: 0.17220
net/net-c2s-cork.js len=16 type=buf dur=5: 0.33850
net/net-c2s-cork.js len=32 type=buf dur=5: 0.61596
net/net-c2s-cork.js len=64 type=buf dur=5: 1.35606
net/net-c2s-cork.js len=128 type=buf dur=5: 2.40356
net/net-c2s-cork.js len=512 type=buf dur=5: 4.86006
net/net-c2s-cork.js len=1024 type=buf dur=5: 5.41967

This were taken on a Macbook Pro 2014 (i7, 16 GB of RAM), and there is quite a variance here. If anybody has a physical linux box (non-virtualized) and wants to test, please do.

This follow up the work on #3751.

cc @chrisdickinson @jasnell @geek @cjihrig @nodejs/streams

mscdex · 2015-12-19T21:25:52Z

I wonder if this would be a good time to create a separate set of benchmarks for streams directly, that way networking, file systems, and the like won't influence results when measuring changes to streams?

ronkorving · 2015-12-20T03:28:19Z

@mscdex +1 on that.

mcollina · 2015-12-20T11:09:12Z

👍 on more benchmarks for streams. However we need to take into account that having real I/O produces some different results anyway.

brendanashworth · 2015-12-31T22:36:34Z

benchmark/net/net-c2s-cork.js

+
+var bench = common.createBenchmark(main, {
+  len: [4, 8, 16, 32, 64, 128, 512, 1024],
+  type: ['buf'],


Is there a reason to keep the other encoding stuff if we only have buffers here?

The reason is because it can be tuned passing arguments. I have disabled them because they do not really provide more data for this benchmark. I can get them removed completely, or re-enable them in the bench.

jasnell · 2016-01-15T01:21:42Z

ping @mcollina ... anything further on this one?

Improves the performance when moving small buffers by 5%, and it adds a benchmark to avoid regression in that area. In all other cases it is equally performant to current master. Full performance results available at: https://gist.github.com/mcollina/717c35ad07d15710b6b9.

mcollina · 2016-02-03T14:03:53Z

This is done for me, I've just rebased and did a final checks. On Linux its more likely to be 8-10% perf benefit with small chunks, and close to no benefit on big chunks. I have tested on Linux using a VM, so somebody running Linux natively should double-check.

jasnell · 2016-02-04T18:07:35Z

LGTM if CI is green: https://ci.nodejs.org/job/node-test-pull-request/1547/

mcollina · 2016-02-04T19:03:55Z

I think this is unrelated: https://ci.nodejs.org/job/node-test-binary-arm/952/RUN_SUBSET=4,nodes=pi2-raspbian-wheezy/tapTestReport/test.tap-138/

jasnell · 2016-02-04T19:05:11Z

Agreed.

mcollina · 2016-02-06T10:55:50Z

Any other opinions on this? @nodejs/streams?

I plan to get this merged in next week.

jasnell · 2016-02-07T21:11:21Z

should be good to go

mcollina · 2016-02-09T09:31:39Z

Landed in 7764b6c

Improves the performance when moving small buffers by 5%, and it adds a benchmark to avoid regression in that area. In all other cases it is equally performant to current master. Full performance results available at: https://gist.github.com/mcollina/717c35ad07d15710b6b9. PR-URL: #4354 Reviewed-By: James M Snell <[email protected]>

* buffer: - You can now supply an encoding argument when filling a Buffer Buffer#fill(string[, start[, end]][, encoding]), supplying an existing Buffer will also work with Buffer#fill(buffer[, start[, end]]). See the API documentation for details on how this works. (Trevor Norris) #4935 - Buffer#indexOf() no longer requires a byteOffset argument if you also wish to specify an encoding: Buffer#indexOf(val[, byteOffset][, encoding]). (Trevor Norris) #4803 * child_process: spawn() and spawnSync() now support a 'shell' option to allow for optional execution of the given command inside a shell. If set to true, cmd.exe will be used on Windows and /bin/sh elsewhere. A path to a custom shell can also be passed to override these defaults. On Windows, this option allows .bat. and .cmd files to be executed with spawn() and spawnSync(). (Colin Ihrig) #4598 * http_parser: Update to http-parser 2.6.2 to fix an unintentionally strict limitation of allowable header characters. (James M Snell) #5237 * dgram: socket.send() now supports accepts an array of Buffers or Strings as the first argument. See the API docs for details on how this works. (Matteo Collina) #4374 * http: Fix a bug where handling headers will mistakenly trigger an 'upgrade' event where the server is just advertising its protocols. This bug can prevent HTTP clients from communicating with HTTP/2 enabled servers. (Fedor Indutny) #4337 * net: Added a listening Boolean property to net and http servers to indicate whether the server is listening for connections. (José Moreira) #4743 * node: The C++ node::MakeCallback() API is now reentrant and calling it from inside another MakeCallback() call no longer causes the nextTick queue or Promises microtask queue to be processed out of order. (Trevor Norris) #4507 * tls: Add a new tlsSocket.getProtocol() method to get the negotiated TLS protocol version of the current connection. (Brian White) #4995 * vm: Introduce new 'produceCachedData' and 'cachedData' options to new vm.Script() to interact with V8's code cache. When a new vm.Script object is created with the 'produceCachedData' set to true a Buffer with V8's code cache data will be produced and stored in cachedData property of the returned object. This data in turn may be supplied back to another vm.Script() object with a 'cachedData' option if the supplied source is the same. Successfully executing a script from cached data can speed up instantiation time. See the API docs for details. (Fedor Indutny) #4777 * performance: Improvements in: - process.nextTick() (Ruben Bridgewater) #5092 - path module (Brian White) #5123 - querystring module (Brian White) #5012 - streams module when processing small chunks (Matteo Collina) #4354

* buffer: - You can now supply an encoding argument when filling a Buffer Buffer#fill(string[, start[, end]][, encoding]), supplying an existing Buffer will also work with Buffer#fill(buffer[, start[, end]]). See the API documentation for details on how this works. (Trevor Norris) #4935 - Buffer#indexOf() no longer requires a byteOffset argument if you also wish to specify an encoding: Buffer#indexOf(val[, byteOffset][, encoding]). (Trevor Norris) #4803 * child_process: spawn() and spawnSync() now support a 'shell' option to allow for optional execution of the given command inside a shell. If set to true, cmd.exe will be used on Windows and /bin/sh elsewhere. A path to a custom shell can also be passed to override these defaults. On Windows, this option allows .bat. and .cmd files to be executed with spawn() and spawnSync(). (Colin Ihrig) #4598 * http_parser: Update to http-parser 2.6.2 to fix an unintentionally strict limitation of allowable header characters. (James M Snell) #5237 * dgram: socket.send() now supports accepts an array of Buffers or Strings as the first argument. See the API docs for details on how this works. (Matteo Collina) #4374 * http: Fix a bug where handling headers will mistakenly trigger an 'upgrade' event where the server is just advertising its protocols. This bug can prevent HTTP clients from communicating with HTTP/2 enabled servers. (Fedor Indutny) #4337 * net: Added a listening Boolean property to net and http servers to indicate whether the server is listening for connections. (José Moreira) #4743 * node: The C++ node::MakeCallback() API is now reentrant and calling it from inside another MakeCallback() call no longer causes the nextTick queue or Promises microtask queue to be processed out of order. (Trevor Norris) #4507 * tls: Add a new tlsSocket.getProtocol() method to get the negotiated TLS protocol version of the current connection. (Brian White) #4995 * vm: Introduce new 'produceCachedData' and 'cachedData' options to new vm.Script() to interact with V8's code cache. When a new vm.Script object is created with the 'produceCachedData' set to true a Buffer with V8's code cache data will be produced and stored in cachedData property of the returned object. This data in turn may be supplied back to another vm.Script() object with a 'cachedData' option if the supplied source is the same. Successfully executing a script from cached data can speed up instantiation time. See the API docs for details. (Fedor Indutny) #4777 * performance: Improvements in: - process.nextTick() (Ruben Bridgewater) #5092 - path module (Brian White) #5123 - querystring module (Brian White) #5012 - streams module when processing small chunks (Matteo Collina) #4354 PR-URL: #5295

reqshark · 2016-02-23T21:37:40Z

strong work!

In December we announced that we would be doing a minor release in order to get a number of voted on SEMVER-MINOR changes into LTS. Our ability to release this was delayed due to the unforeseen security release v4.3. We are quickly bumping to v4.4 in order to bring you the features that we had committed to releasing. This release also includes security updates to openssl. More information can be found [on nodejs.org](https://nodejs.org/en/blog/vulnerability/openssl-march-2016/) This release also includes over 70 fixes to our docs and over 50 fixes to tests. The SEMVER-MINOR changes include: * deps: - An update to v8 that introduces a new flag --perf_basic_prof_only_functions (Ali Ijaz Sheikh) #3609 * http: - A new feature in http(s) agent that catches errors on *keep alived* connections (José F. Romaniello) #4482 * src: - Better support for Big-Endian systems (Bryon Leung) #3410 * tls: - A new feature that allows you to pass common SSL options to `tls.createSecurePair` (Коренберг Марк) #2441 * tools - a new flag `--prof-process` which will execute the tick processor on the provided isolate files (Matt Loring) #4021 Notable semver patch changes include: * buld: - Support python path that includes spaces. This should be of particular interest to our Windows users who may have python living in `c:/Program Files` (Felix Becker) #4841 * https: - A potential fix for #3692 HTTP/HTTPS client requests throwing EPROTO (Fedor Indutny) #4982 * installer: - More readable profiling information from isolate tick logs (Matt Loring) #3032 * *npm: - upgrade to npm 2.14.20 (Kat Marchán) #5510 * *openssl: - upgrade openssl to 1.0.2g (Ben Noordhuis) #5507 * process: - Add support for symbols in event emitters. Symbols didn't exist when it was written ¯\_(ツ)_/¯ (cjihrig) #4798 * querystring: - querystring.parse() is now 13-22% faster! (Brian White) #4675 * streams: - performance improvements for moving small buffers that shows a 5% throughput gain. IoT projects have been seen to be as much as 10% faster with this change! (Matteo Collina) #4354 * tools: - eslint has been updated to version 2.1.0 (Rich Trott) #5214 PR-URL: #5301

In December we announced that we would be doing a minor release in order to get a number of voted on SEMVER-MINOR changes into LTS. Our ability to release this was delayed due to the unforeseen security release v4.3. We are quickly bumping to v4.4 in order to bring you the features that we had committed to releasing. This release also includes security updates to openssl. More information can be found [on nodejs.org](https://nodejs.org/en/blog/vulnerability/openssl-march-2016/) This release also includes over 70 fixes to our docs and over 50 fixes to tests. The SEMVER-MINOR changes include: * deps: - An update to v8 that introduces a new flag --perf_basic_prof_only_functions (Ali Ijaz Sheikh) nodejs#3609 * http: - A new feature in http(s) agent that catches errors on *keep alived* connections (José F. Romaniello) nodejs#4482 * src: - Better support for Big-Endian systems (Bryon Leung) nodejs#3410 * tls: - A new feature that allows you to pass common SSL options to `tls.createSecurePair` (Коренберг Марк) nodejs#2441 * tools - a new flag `--prof-process` which will execute the tick processor on the provided isolate files (Matt Loring) nodejs#4021 Notable semver patch changes include: * buld: - Support python path that includes spaces. This should be of particular interest to our Windows users who may have python living in `c:/Program Files` (Felix Becker) nodejs#4841 * https: - A potential fix for nodejs#3692 HTTP/HTTPS client requests throwing EPROTO (Fedor Indutny) nodejs#4982 * installer: - More readable profiling information from isolate tick logs (Matt Loring) nodejs#3032 * *npm: - upgrade to npm 2.14.20 (Kat Marchán) nodejs#5510 * *openssl: - upgrade openssl to 1.0.2g (Ben Noordhuis) nodejs#5507 * process: - Add support for symbols in event emitters. Symbols didn't exist when it was written ¯\_(ツ)_/¯ (cjihrig) nodejs#4798 * querystring: - querystring.parse() is now 13-22% faster! (Brian White) nodejs#4675 * streams: - performance improvements for moving small buffers that shows a 5% throughput gain. IoT projects have been seen to be as much as 10% faster with this change! (Matteo Collina) nodejs#4354 * tools: - eslint has been updated to version 2.1.0 (Rich Trott) nodejs#5214 PR-URL: nodejs#5301

Improves the performance when moving small buffers by 5%, and it adds a benchmark to avoid regression in that area. In all other cases it is equally performant to current master. Full performance results available at: https://gist.github.com/mcollina/717c35ad07d15710b6b9. PR-URL: #4354 Reviewed-By: James M Snell <[email protected]>

In December we announced that we would be doing a minor release in order to get a number of voted on SEMVER-MINOR changes into LTS. Our ability to release this was delayed due to the unforeseen security release v4.3. We are quickly bumping to v4.4 in order to bring you the features that we had committed to releasing. This release also includes over 70 fixes to our docs and over 50 fixes to tests. The SEMVER-MINOR changes include: * deps: - An update to v8 that introduces a new flag --perf_basic_prof_only_functions (Ali Ijaz Sheikh) #3609 * http: - A new feature in http(s) agent that catches errors on *keep alived* connections (José F. Romaniello) #4482 * src: - Better support for Big-Endian systems (Bryon Leung) #3410 * tls: - A new feature that allows you to pass common SSL options to `tls.createSecurePair` (Коренберг Марк) #2441 * tools - a new flag `--prof-process` which will execute the tick processor on the provided isolate files (Matt Loring) #4021 Notable semver patch changes include: * buld: - Support python path that includes spaces. This should be of particular interest to our Windows users who may have python living in `c:/Program Files` (Felix Becker) #4841 * https: - A potential fix for #3692 HTTP/HTTPS client requests throwing EPROTO (Fedor Indutny) #4982 * installer: - More readable profiling information from isolate tick logs (Matt Loring) #3032 * *npm: - upgrade to npm 2.14.20 (Kat Marchán) #5510 * process: - Add support for symbols in event emitters. Symbols didn't exist when it was written ¯\_(ツ)_/¯ (cjihrig) #4798 * querystring: - querystring.parse() is now 13-22% faster! (Brian White) #4675 * streams: - performance improvements for moving small buffers that shows a 5% throughput gain. IoT projects have been seen to be as much as 10% faster with this change! (Matteo Collina) #4354 * tools: - eslint has been updated to version 2.1.0 (Rich Trott) #5214 PR-URL: #5301

net streams can request multiple chunks to be written in a synchronous fashion. If this is combined with cork/uncork, en error is currently thrown because of a regression introduced in: nodejs@89aeab9 (nodejs#4354). Fixes: nodejs#6154.

net streams can request multiple chunks to be written in a synchronous fashion. If this is combined with cork/uncork, en error is currently thrown because of a regression introduced in: 89aeab9 (#4354). Fixes: #6154 PR-URL: #6164 Reviewed-By: Benjamin Gruenbaum <[email protected]> Reviewed-By: Mathias Buus <[email protected]> Reviewed-By: James M Snell <[email protected]>

mscdex added the stream Issues and PRs related to the stream subsystem. label Dec 19, 2015

jasnell added the lts-watch-v4.x label Dec 24, 2015

Trott force-pushed the master branch from 1e896a6 to 082cc8d Compare December 27, 2015 02:01

brendanashworth reviewed Dec 31, 2015
View reviewed changes

mcollina force-pushed the clearBuffer-small-chunks branch from 3e6220c to f990d62 Compare February 3, 2016 12:06

mcollina closed this Feb 9, 2016

mcollina deleted the clearBuffer-small-chunks branch February 9, 2016 09:31

rvagg mentioned this pull request Feb 18, 2016

Release proposal: 5.7.0 (Stable) #5295

Merged

ChALkeR added the performance Issues and PRs related to the performance of Node.js. label Feb 23, 2016

MylesBorins mentioned this pull request Mar 1, 2016

V4.4.0 proposal #5301

Merged

mcollina mentioned this pull request Apr 12, 2016

streams: support unlimited synchronous cork/uncork cycles #6164

Closed

2 tasks

mcollina mentioned this pull request May 30, 2016

http: server code refactor #6533

Merged

2 tasks

snyk-bot mentioned this pull request Feb 12, 2022

[Snyk] Upgrade rollup from 2.58.0 to 2.65.0 Iceymann18777/node#12

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

streams: 5% throughput gain when sending small chunks #4354

streams: 5% throughput gain when sending small chunks #4354

mcollina commented Dec 19, 2015

mscdex commented Dec 19, 2015

ronkorving commented Dec 20, 2015

mcollina commented Dec 20, 2015

brendanashworth Dec 31, 2015

mcollina Feb 4, 2016

jasnell commented Jan 15, 2016

mcollina commented Feb 3, 2016

jasnell commented Feb 4, 2016

mcollina commented Feb 4, 2016

jasnell commented Feb 4, 2016

mcollina commented Feb 6, 2016

jasnell commented Feb 7, 2016

mcollina commented Feb 9, 2016

reqshark commented Feb 23, 2016

streams: 5% throughput gain when sending small chunks #4354

streams: 5% throughput gain when sending small chunks #4354

Conversation

mcollina commented Dec 19, 2015

mscdex commented Dec 19, 2015

ronkorving commented Dec 20, 2015

mcollina commented Dec 20, 2015

brendanashworth Dec 31, 2015

Choose a reason for hiding this comment

mcollina Feb 4, 2016

Choose a reason for hiding this comment

jasnell commented Jan 15, 2016

mcollina commented Feb 3, 2016

jasnell commented Feb 4, 2016

mcollina commented Feb 4, 2016

jasnell commented Feb 4, 2016

mcollina commented Feb 6, 2016

jasnell commented Feb 7, 2016

mcollina commented Feb 9, 2016

reqshark commented Feb 23, 2016