code review 7300075: net/http: more request benchmarks

net/http: more request benchmarks

Add benchmarks for common http benchmarking tools. The intent is to catch optimisations which favor synthetic benchmarks that do not show improvements for real clients like Chrome.

BenchmarkReadRequestChrome        200000             10133 ns/op          60.29 MB/s        3148 B/op         32 allocs/op
BenchmarkReadRequestCurl          500000              4314 ns/op          18.08 MB/s         905 B/op         15 allocs/op
BenchmarkReadRequestApachebench   500000              4363 ns/op          18.79 MB/s         956 B/op         16 allocs/op
BenchmarkReadRequestSiege         500000              6408 ns/op          24.19 MB/s        1397 B/op         22 allocs/op
BenchmarkReadRequestWrk          1000000              2838 ns/op          14.09 MB/s         757 B/op         11 allocs/op

[dave_cheney.net, 2013-02-09 00:31:09]

Hello golang-dev@googlegroups.com (cc: haimuiba@gmail.com),

I'd like you to review this change to
https://code.google.com/p/go

[dave_cheney.net, 2013-02-09 00:37:49]

To give some back story here, most of the benchmarking tools appear to use
requests that are much smaller and simpler than a real world example. Wrk is the
worst, with only a Host: header.

Consider this, https://codereview.appspot.com/7306073/ which shows an 8%
improvement in parsing for Chrome, but 19% regression for Wrk.

[bradfitz, 2013-02-09 01:14:09]

LGTM

Sure.  I'd drop the "const request" and inline the string literal. And drop
the double \n\n suffix from them and add it in the helper.

On Fri, Feb 8, 2013 at 4:31 PM, <dave@cheney.net> wrote:

> Reviewers: golang-dev_googlegroups.com,
>
> Message:
> Hello golang-dev@googlegroups.com (cc: haimuiba@gmail.com),
>
> I'd like you to review this change to
> https://code.google.com/p/go
>
>
> Description:
> net/http: more request benchmarks
>
> Add benchmarks for common http benchmarking tools. The intent is to
> catch optimisations which favor synthetic benchmarks that do not show
> improvements for real clients like Chrome.
>
> BenchmarkReadRequestChrome        200000             10133 ns/op
>  60.29 MB/s        3148 B/op         32 allocs/op
> BenchmarkReadRequestCurl          500000              4314 ns/op
>  18.08 MB/s         905 B/op         15 allocs/op
> BenchmarkReadRequestApachebenc**h   500000              4363 ns/op
>  18.79 MB/s         956 B/op         16 allocs/op
> BenchmarkReadRequestSiege         500000              6408 ns/op
>  24.19 MB/s        1397 B/op         22 allocs/op
> BenchmarkReadRequestWrk          1000000              2838 ns/op
>  14.09 MB/s         757 B/op         11 allocs/op
>
> Please review this at
https://codereview.appspot.**com/7300075/<https://codereview.appspot.com/7300...
>
> Affected files:
>   M src/pkg/net/http/request_test.**go
>
>
> Index: src/pkg/net/http/request_test.**go
> ==============================**==============================**=======
> --- a/src/pkg/net/http/request_**test.go
> +++ b/src/pkg/net/http/request_**test.go
> @@ -403,6 +403,7 @@
>  `
>
>  func benchmarkReadRequest(b *testing.B, request string) {
> +       request = strings.Replace(request, "\n", "\r\n", -1)
>         b.SetBytes(int64(len(request))**)
>         r := bufio.NewReader(&**infiniteReader{buf: []byte(request)})
>         b.ReportAllocs()
> @@ -435,7 +436,7 @@
>         return a
>  }
>
> -func BenchmarkReadRequest(b *testing.B) {
> +func BenchmarkReadRequestChrome(b *testing.B) {
>         // https://github.com/felixge/**node-http-perf/blob/master/**
>
fixtures/get.http<https://github.com/felixge/node-http-perf/blob/master/fixtures/get.http>
>         const request = `GET / HTTP/1.1
>  Host: localhost:8080
> @@ -448,5 +449,49 @@
>  Cookie: __utma=1.1978842379.**1323102373.1323102373.**1323102373.1;
> EPi:NumberOfVisits=1,2012-02-**28T13:42:18; CrmSession=**
> 5b707226b9563e1bc69084d07a107c**98; plushContainerWidth=100%25;
> plushNoTopMenu=0; hudson_auto_refresh=false
>
>  `
> -       benchmarkReadRequest(b, strings.Replace(request, "\n", "\r\n", -1))
> +       benchmarkReadRequest(b, request)
>  }
> +
> +func BenchmarkReadRequestCurl(b *testing.B) {
> +       // curl http://localhost:8080/
> +       const request = `GET / HTTP/1.1
> +User-Agent: curl/7.27.0
> +Host: localhost:8080
> +Accept: */*
> +
> +`
> +       benchmarkReadRequest(b, request)
> +}
> +
> +func BenchmarkReadRequestApachebenc**h(b *testing.B) {
> +       // ab -n 1 -c 1 http://localhost:8080/
> +       const request = `GET / HTTP/1.0
> +Host: localhost:8080
> +User-Agent: ApacheBench/2.3
> +Accept: */*
> +
> +`
> +       benchmarkReadRequest(b, request)
> +}
> +
> +func BenchmarkReadRequestSiege(b *testing.B) {
> +       // siege -r 1 -c 1 http://localhost:8080/
> +       const request = `GET / HTTP/1.1
> +Host: localhost:8080
> +Accept: */*
> +Accept-Encoding: gzip
> +User-Agent: JoeDog/1.00 [en] (X11; I; Siege 2.70)
> +Connection: keep-alive....
> +
> +`
> +       benchmarkReadRequest(b, request)
> +}
> +
> +func BenchmarkReadRequestWrk(b *testing.B) {
> +       // wrk -t 1 -r 1 -c 1 http://localhost:8080/
> +       const request = `GET / HTTP/1.1
> +Host: localhost:8080
> +
> +`
> +       benchmarkReadRequest(b, request)
> +}
>
>
> --
>
> ---You received this message because you are subscribed to the Google
> Groups "golang-dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to
golang-dev+unsubscribe@**googlegroups.com<golang-dev%2Bunsubscribe@googlegrou...
> .
> For more options, visit
https://groups.google.com/**groups/opt_out<https://groups.google.com/groups/o...
> .
>
>
>

[dave_cheney.net, 2013-02-09 07:55:58]

Hello golang-dev@googlegroups.com, bradfitz@golang.org (cc:
golang-dev@googlegroups.com, haimuiba@gmail.com),

Please take another look.

[dave_cheney.net, 2013-02-09 21:18:25]

*** Submitted as https://code.google.com/p/go/source/detail?r=5c213525c4b5 ***

net/http: more request benchmarks

Add benchmarks for common http benchmarking tools. The intent is to catch
optimisations which favor synthetic benchmarks that do not show improvements for
real clients like Chrome.

BenchmarkReadRequestChrome        200000             10133 ns/op          60.29
MB/s        3148 B/op         32 allocs/op
BenchmarkReadRequestCurl          500000              4314 ns/op          18.08
MB/s         905 B/op         15 allocs/op
BenchmarkReadRequestApachebench   500000              4363 ns/op          18.79
MB/s         956 B/op         16 allocs/op
BenchmarkReadRequestSiege         500000              6408 ns/op          24.19
MB/s        1397 B/op         22 allocs/op
BenchmarkReadRequestWrk          1000000              2838 ns/op          14.09
MB/s         757 B/op         11 allocs/op

R=golang-dev, bradfitz
CC=golang-dev, haimuiba
https://codereview.appspot.com/7300075