[chore][confighttp] use sync.Pool

[florianl]

Description

When profiling some environments excessive calls to bytes.NewBuffer() and consequently additional GC workload was noticed. Use sync.Pool to reduce this overhead.

Link to tracking issue

Fixes #

Testing

Documentation

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.63%. Comparing base (d00f059) to head (e53af78).

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #13909      +/-   ##
==========================================
- Coverage   91.65%   91.63%   -0.03%     
==========================================
  Files         652      652              
  Lines       42516    42523       +7     
==========================================
- Hits        38970    38965       -5     
- Misses       2736     2745       +9     
- Partials      810      813       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[dmathieu]

Could you share either benchmarks showing the improvement, or profiles before and after this change?

[florianl]

@dmathieu here are some numbers:

$ go test -bench=BenchmarkCompressRoundTripper -run=^$ -benchtime=8s
goos: linux
goarch: amd64
pkg: go.opentelemetry.io/collector/config/confighttp
cpu: 12th Gen Intel(R) Core(TM) i7-12700H
BenchmarkCompressRoundTripper/Small_1KB_gzip_WithPool-20                  557988             26724 ns/op           28703 B/op        112 allocs/op
BenchmarkCompressRoundTripper/Small_1KB_gzip_NoPool-20                    570410             27245 ns/op           30017 B/op        115 allocs/op
BenchmarkCompressRoundTripper/Small_1KB_zstd_WithPool-20                  725426             21008 ns/op           18391 B/op        107 allocs/op
BenchmarkCompressRoundTripper/Small_1KB_zstd_NoPool-20                    763928             12327 ns/op           19925 B/op        111 allocs/op
BenchmarkCompressRoundTripper/Small_1KB_snappy_WithPool-20                281094             33449 ns/op           26842 B/op        112 allocs/op
BenchmarkCompressRoundTripper/Small_1KB_snappy_NoPool-20                  295188             35492 ns/op           27938 B/op        114 allocs/op
BenchmarkCompressRoundTripper/Medium_10KB_gzip_WithPool-20                438165             25951 ns/op           26833 B/op        113 allocs/op
BenchmarkCompressRoundTripper/Medium_10KB_gzip_NoPool-20                  405530             26554 ns/op           28015 B/op        117 allocs/op
BenchmarkCompressRoundTripper/Medium_10KB_zstd_WithPool-20                860119             21738 ns/op           20865 B/op        108 allocs/op
BenchmarkCompressRoundTripper/Medium_10KB_zstd_NoPool-20                  639778             22276 ns/op           21169 B/op        110 allocs/op
BenchmarkCompressRoundTripper/Medium_10KB_snappy_WithPool-20              261571             36828 ns/op           43490 B/op        113 allocs/op
BenchmarkCompressRoundTripper/Medium_10KB_snappy_NoPool-20                252706             38589 ns/op           44842 B/op        115 allocs/op
BenchmarkCompressRoundTripper/Large_100KB_gzip_WithPool-20                195235             46352 ns/op           16755 B/op        107 allocs/op
BenchmarkCompressRoundTripper/Large_100KB_gzip_NoPool-20                  186924             45780 ns/op           18505 B/op        112 allocs/op
BenchmarkCompressRoundTripper/Large_100KB_zstd_WithPool-20                596923             23384 ns/op           21277 B/op        111 allocs/op
BenchmarkCompressRoundTripper/Large_100KB_zstd_NoPool-20                  616399             23776 ns/op           23305 B/op        115 allocs/op
BenchmarkCompressRoundTripper/Large_100KB_snappy_WithPool-20              162502             55986 ns/op          202809 B/op        122 allocs/op
BenchmarkCompressRoundTripper/Large_100KB_snappy_NoPool-20                163826             50143 ns/op          198925 B/op        123 allocs/op
BenchmarkCompressRoundTripper/XLarge_1MB_gzip_WithPool-20                  32559            260401 ns/op            8576 B/op         87 allocs/op
BenchmarkCompressRoundTripper/XLarge_1MB_gzip_NoPool-20                    35058            263069 ns/op           26055 B/op         96 allocs/op
BenchmarkCompressRoundTripper/XLarge_1MB_zstd_WithPool-20                 273133             36299 ns/op           18345 B/op        112 allocs/op
BenchmarkCompressRoundTripper/XLarge_1MB_zstd_NoPool-20                   270033             36723 ns/op           18704 B/op        116 allocs/op
BenchmarkCompressRoundTripper/XLarge_1MB_snappy_WithPool-20                59028            154116 ns/op         1493545 B/op        123 allocs/op
BenchmarkCompressRoundTripper/XLarge_1MB_snappy_NoPool-20                  62806            152046 ns/op         1497548 B/op        124 allocs/op
PASS
ok      go.opentelemetry.io/collector/config/confighttp 316.170s

Feel free to reproduce the results with https://gist.github.com/florianl/22b50787ce38f0d637df17ed7875be42.

The benchmark is limited as they show only a limited number of http requests. When applying this change with more http requests and for a longer time, the benefits will be increase.

[atoulme]

sorry @florianl would you please offer the benchmark results before/after on your computer so we can see the diff, using a tool like benchstat?