iris/_benchmarks/_internal/README.md

Benchmarks (internal)

Internal selected benchmarks between modified features across different versions.

• These benchmarks SHOULD run locally with a break of ~2 minutes between stress tests at the same machine, power plan and Turbo Boost set to ON
• The system's GOPATH environment variable SHOULD match the vNext/go.mod replace directive one
• A stress test may ran from a name_test.go file to measure BUILD TIME
  • each version executes: go test -run=NONE --bench=. -count=5 --benchmem > name_test.txt
  • the result will be presented through benchstat tool
• Or/and by firing bombardier HTTP requests when it (the test) listens to an address
• Each benchmark SHOULD contain a brief explanation of what it does.

Dependency Injection

Measures handler factory time.

Name	Ops	Ns/op	B/op	Allocs/op
vNext	181726	6631	1544	17
v12.1.x	96001	12604	976	26

It accepts a dynamic path parameter and a JSON request. It returns a JSON response. Fires 500000 requests with 125 concurrent connections.

# di.go
$ bombardier -c 125 -n 500000 --method="POST" --body-file=./request.json http://localhost:5000/42

Name	Throughput	Reqs/sec	Latency	Time To Complete
vNext	46.51MB/s	160480.74	777.33us	3s
v12.1.x	32.43MB/s	108839.19	1.14ms	4s