wrk는 HTTP 벤치마크 툴이다.
$ wrk -t7 -c100 -d10s https://google.com
Running 10s test @ https://google.com
7 threads and 100 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 77.10ms 7.98ms 251.74ms 92.75%
Req/Sec 181.05 23.21 240.00 82.83%
12542 requests in 10.08s, 8.77MB read
Requests/sec: 1243.90
Transfer/sec: 0.87MB
위 명령어를 입력하면 쓰레드 7개, 최대 100개의 동시 연결로 10초동안 구글에 요청을 마구 보낸다.
wrk의 latency 측정 방식에 문제가 좀 있는지 누군가가 개선하여 wrk2라는 것도 내놓았다.
$ wrk2 -R1000 -t7 -c100 -d10s https://google.com
Running 10s test @ https://google.com
7 threads and 100 connections
Thread calibration: mean lat.: 4156.029ms, rate sampling interval: 14344ms
Thread Stats Avg Stdev Max +/- Stdev
Latency 4.10s 2.09s 7.97s 57.80%
Req/Sec -nan -nan 0.00 0.00%
2135 requests in 10.07s, 1.92MB read
Requests/sec: 211.99
Transfer/sec: 195.02KB위 명령어를 입력하면 쓰레드 7개, 최대 100개의 동시 연결로 초당 1000개의 요청을 10초 동안 구글에 날린다.
-L 또는 --latency 옵션을 주면 더 자세한 레이턴시 결과를 얻을 수 있다.
$ wrk2 -R1000 -t7 -c100 -d10s -L https://google.com
Running 10s test @ https://google.com
7 threads and 100 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 3.96s 2.07s 7.96s 58.14%
Req/Sec -nan -nan 0.00 0.00%
Latency Distribution (HdrHistogram - Recorded Latency)
50.000% 3.95s
75.000% 5.72s
90.000% 6.79s
99.000% 7.73s
99.900% 7.93s
99.990% 7.97s
99.999% 7.97s
100.000% 7.97s
Detailed Percentile spectrum:
Value Percentile TotalCount 1/(1-Percentile)
391.679 0.000000 1 1.00
1116.159 0.100000 212 1.11
1807.359 0.200000 422 1.25
2545.663 0.300000 633 1.43
3235.839 0.400000 843 1.67
3946.495 0.500000 1054 2.00
4304.895 0.550000 1160 2.22
4661.247 0.600000 1266 2.50
5005.311 0.650000 1371 2.86
5349.375 0.700000 1475 3.33
5718.015 0.750000 1581 4.00
5881.855 0.775000 1633 4.44
6086.655 0.800000 1688 5.00
6254.591 0.825000 1740 5.71
6426.623 0.850000 1791 6.67
6619.135 0.875000 1844 8.00
6713.343 0.887500 1870 8.89
6787.071 0.900000 1897 10.00
6868.991 0.912500 1925 11.43
6983.679 0.925000 1950 13.33
7090.175 0.937500 1977 16.00
7135.231 0.943750 1989 17.78
7221.247 0.950000 2002 20.00
7315.455 0.956250 2015 22.86
7364.607 0.962500 2028 26.67
7458.815 0.968750 2042 32.00
7512.063 0.971875 2049 35.56
7536.639 0.975000 2055 40.00
7557.119 0.978125 2061 45.71
7618.559 0.981250 2068 53.33
7679.999 0.984375 2076 64.00
7684.095 0.985938 2078 71.11
7696.383 0.987500 2081 80.00
7704.575 0.989062 2085 91.43
7757.823 0.990625 2088 106.67
7766.015 0.992188 2091 128.00
7774.207 0.992969 2094 142.22
7774.207 0.993750 2094 160.00
7794.687 0.994531 2097 182.86
7802.879 0.995313 2098 213.33
7811.071 0.996094 2099 256.00
7823.359 0.996484 2100 284.44
7860.223 0.996875 2101 320.00
7876.607 0.997266 2102 365.71
7880.703 0.997656 2103 426.67
7880.703 0.998047 2103 512.00
7909.375 0.998242 2104 568.89
7909.375 0.998437 2104 640.00
7933.951 0.998633 2105 731.43
7933.951 0.998828 2105 853.33
7933.951 0.999023 2105 1024.00
7946.239 0.999121 2106 1137.78
7946.239 0.999219 2106 1280.00
7946.239 0.999316 2106 1462.86
7946.239 0.999414 2106 1706.67
7946.239 0.999512 2106 2048.00
7966.719 0.999561 2107 2275.56
7966.719 1.000000 2107 inf
#[Mean = 3963.857, StdDeviation = 2066.100]
#[Max = 7962.624, Total count = 2107]
#[Buckets = 27, SubBuckets = 2048]
----------------------------------------------------------
2114 requests in 10.02s, 1.90MB read
Requests/sec: 210.93
Transfer/sec: 194.04KB위 데이터를 http://hdrhistogram.org/ 에 업로드하면 레이턴시 분포를 그려준다.