Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Reduce slice growth in adaptor #1356

Merged
merged 1 commit into from Aug 16, 2022

Conversation

moycat
Copy link
Contributor

@moycat moycat commented Aug 15, 2022

This PR improves the performance of fasthttpadaptor by reducing the overhead of slice growth.

The current adapter records all written body in a byte slice, which causes horrible overhead when the body is bigger than a few kilobytes. However, this is unnecessary as we can just pass the response writer of fasthttp.

func (w *netHTTPResponseWriter) Write(p []byte) (int, error) {
w.body = append(w.body, p...)
return len(p), nil
}

I compared the performance by modifying the ServerGet tests with 32 KiB responses.

The original ServerGet without the adaptor:

BenchmarkServerGet1ReqPerConn-4                          4658205              2622 ns/op               0 B/op          0 allocs/op
BenchmarkServerGet2ReqPerConn-4                          5202049              2466 ns/op               0 B/op          0 allocs/op
BenchmarkServerGet10ReqPerConn-4                         7114965              1754 ns/op               0 B/op          0 allocs/op
BenchmarkServerGet10KReqPerConn-4                        8304604              1486 ns/op               0 B/op          0 allocs/op
BenchmarkServerGet1ReqPerConn10KClients-4                4690203              2395 ns/op              12 B/op          0 allocs/op
BenchmarkServerGet2ReqPerConn10KClients-4                5133776              2260 ns/op               3 B/op          0 allocs/op
BenchmarkServerGet10ReqPerConn10KClients-4               7322839              1583 ns/op               1 B/op          0 allocs/op
BenchmarkServerGet100ReqPerConn10KClients-4              8498665              1402 ns/op               2 B/op          0 allocs/op

The ServerGet with the current adaptor:

BenchmarkAdaptorGet1ReqPerConn-4                 1000000             14145 ns/op           36900 B/op         33 allocs/op
BenchmarkAdaptorGet2ReqPerConn-4                  819870             13954 ns/op           36749 B/op         32 allocs/op
BenchmarkAdaptorGet10ReqPerConn-4                 932516             11769 ns/op           35295 B/op         30 allocs/op
BenchmarkAdaptorGet10KReqPerConn-4               1116939             10816 ns/op           34562 B/op         30 allocs/op
BenchmarkAdaptorGet1ReqPerConn10KClients-4        558528             24187 ns/op           46985 B/op         44 allocs/op
BenchmarkAdaptorGet2ReqPerConn10KClients-4        745404             14326 ns/op           40126 B/op         35 allocs/op
BenchmarkAdaptorGet10ReqPerConn10KClients-4       920809             10949 ns/op           36740 B/op         32 allocs/op
BenchmarkAdaptorGet100ReqPerConn10KClients-4     1320006              8900 ns/op           34945 B/op         30 allocs/op

The ServerGet with the patched adaptor:

BenchmarkAdaptorGet1ReqPerConn-4                 3693327              3372 ns/op            1911 B/op         30 allocs/op
BenchmarkAdaptorGet2ReqPerConn-4                 3534920              3360 ns/op            1887 B/op         29 allocs/op
BenchmarkAdaptorGet10ReqPerConn-4                4373206              2716 ns/op            1893 B/op         29 allocs/op
BenchmarkAdaptorGet10KReqPerConn-4               4955851              2424 ns/op            1776 B/op         29 allocs/op
BenchmarkAdaptorGet1ReqPerConn10KClients-4       2716551              4831 ns/op            3189 B/op         31 allocs/op
BenchmarkAdaptorGet2ReqPerConn10KClients-4       3118240              4085 ns/op            2953 B/op         30 allocs/op
BenchmarkAdaptorGet10ReqPerConn10KClients-4      4207117              3088 ns/op            2383 B/op         29 allocs/op
BenchmarkAdaptorGet100ReqPerConn10KClients-4     5002465              2379 ns/op            1926 B/op         29 allocs/op

@erikdubbelboer erikdubbelboer merged commit af94725 into valyala:master Aug 16, 2022
@erikdubbelboer
Copy link
Collaborator

Thanks!

bbenzikry pushed a commit to bbenzikry/fasthttp that referenced this pull request Sep 11, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants