Question

我正在运行2种不同实现的wrk基准测试（请参见下面的结果）。

我不明白较低延迟的实现（“ OLD”）如何每秒获得更少的请求。

旧版本的平均延迟时间为150ms，每秒可接收21个请求，而新版本的平均延迟时间为390ms和每秒49个请求。

我期望相反：更低的延迟可以每秒获得更多请求...

“旧”：

wrk -t5 -c20 -d10s --timeout 1s  --latency https://old
Running 10s test @ https://old
  5 threads and 20 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   152.47ms   89.32ms 757.56ms   94.81%
    Req/Sec     7.76      2.78    10.00     70.67%
  Latency Distribution
     50%  133.56ms
     75%  153.22ms
     90%  170.08ms
     99%  727.16ms
  212 requests in 10.08s, 9.05MB read
Requests/sec:     21.04
Transfer/sec:      0.90MB

“新”：

wrk -t5 -c20 -d10s --latency --timeout 1s  https://new
Running 10s test @ https://new
  5 threads and 20 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   389.86ms   74.52ms 618.33ms   70.62%
    Req/Sec    11.65      6.86    30.00     61.25%
  Latency Distribution
     50%  375.55ms
     75%  431.85ms
     90%  493.65ms
     99%  578.83ms
  497 requests in 10.08s, 21.93MB read
Requests/sec:     49.30
Transfer/sec:      2.17MB

wrk中较低的延迟和较低的请求/秒

0 个答案: