use http.ResponseWriter Write() to take payload as http response,the http server's QPS is very low

[he...@ablecloud.cn]

when I use http.ResponseWriter Write() to take payload as http response, in boom test the http server's QPS is just 7000~8000,but if I only set header, in boom test the QPS is up to 40000. is the http.ResponseWriter Write() is enefficiency?

here is my test code:

func handleHeader(w http.ResponseWriter, r *http.Request) {
    w.Header().Set("Content-Length", "0")
}

func handlePayload50(w http.ResponseWriter, r *http.Request) {
   w.Header().Set("Content-Length", "50")
   w.Write(make([]byte, 50))
}

func handlePayload128(w http.ResponseWriter, r *http.Request) {
   w.Header().Set("Content-Length", "128")
   w.Write(make([]byte, 128))
}

func handlePayload1024(w http.ResponseWriter, r *http.Request) {
   w.Header().Set("Content-Length", "1024")
   w.Write(make([]byte, 1024))
}

func main() {
   runtime.GOMAXPROCS(runtime.NumCPU())
   http.HandleFunc("/header", handleHeader)
   http.HandleFunc("/payload50", handlePayload50)
   http.HandleFunc("/payload128", handlePayload128)
   http.HandleFunc("/payload1024", handlePayload1024)
   host := fmt.Sprintf("%s:%d", "0.0.0.0", 5001)
   http.ListenAndServe(host, nil)
}

and the boom test result:

./boom -c 1000 -n 500000 -cpus 12 http://192.168.0.2:5001/header

500000 / 500000 Boooooooooooooooooooooooooooooooooooooooooooooooooooo! 100.00 % 

Summary:

  Total: 12.8799 secs.

  Slowest: 3.2799 secs.

  Fastest: 0.0018 secs.

  Average: 0.0239 secs.

  Requests/sec: 38810.4907

Status code distribution:

  [200] 499875 responses

Response time histogram:

  0.002 [1] |

  0.330 [499306] |∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎

  0.657 [0] |

  0.985 [0] |

  1.313 [312] |

  1.641 [0] |

  1.969 [0] |

  2.297 [0] |

  2.624 [0] |

  2.952 [0] |

  3.280 [256] |

Latency distribution:

  10% in 0.0125 secs.

  25% in 0.0172 secs.

  50% in 0.0216 secs.

  75% in 0.0257 secs.

  90% in 0.0295 secs.

  95% in 0.0327 secs.

  99% in 0.0549 secs.

Error distribution:

  [125] Get http://192.168.0.2:5001/header: dial tcp 192.168.0.2:5001: connection timed out

./boom -c 500 -n 500000 -cpus 12 http://192.168.0.2:5001/payload128

500000 / 500000 Boooooooooooooooooooooooooooooooooooooooooooooooooooo! 100.00 % 

Summary:

  Total: 62.5205 secs.

  Slowest: 3.0714 secs.

  Fastest: 0.0112 secs.

  Average: 0.0601 secs.

  Requests/sec: 7997.1501

  Total Data Received: 63998208 bytes.

  Response Size per Request: 128 bytes.

Status code distribution:

  [200] 499986 responses

Response time histogram:

  0.011 [1] |

  0.317 [495652] |∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎

  0.623 [0] |

  0.929 [0] |

  1.235 [4092] |

  1.541 [0] |

  1.847 [0] |

  2.153 [0] |

  2.459 [0] |

  2.765 [0] |

  3.071 [241] |

Latency distribution:

  10% in 0.0405 secs.

  25% in 0.0436 secs.

  50% in 0.0481 secs.

  75% in 0.0546 secs.

  90% in 0.0629 secs.

  95% in 0.0731 secs.

  99% in 0.1054 secs.

Error distribution:

  [14] Get http://192.168.0.2:5001/payload128: dial tcp 192.168.0.2:5001: connection timed out

[thwd]

Hard to tell from the results alone. Profiling helps:

https://github.com/pkg/profile

- Tom

[he...@ablecloud.cn]

I use the cpu profile, the server cost cpu is low, the new test result come from my pc(the old test result is come from vm), the QPS is up to 14000.the virtual network card is the problem, but I want to know 14000 is the limit in one PC