☰

本地主机上的 http 延迟差异很大

tag-icon • tag-icon linux performance ping real-time latency
本地主机上的 http 延迟差异很大

在具有 10 个核心的 MBP M1 上运行。运行 1 个线程客户端 (wrk) 和 1 个线程服务器 (java)。系统上还运行着 YouTube。平均负载为 2 分（满分 10 分）。除了 GC，以下哪些可能是造成延迟差异巨大的原因？

            "percentile": 0,
            "latency_in_microseconds": 10
        },
        {
            "percentile": 50,
            "latency_in_microseconds": 16
        },
        {
            "percentile": 99,
            "latency_in_microseconds": 51
        },
        {
            "percentile": 99.9,
            "latency_in_microseconds": 1500
        },
        {
            "percentile": 99.99,
            "latency_in_microseconds": 9598
        },
        {
            "percentile": 99.999,
            "latency_in_microseconds": 15021
        }

使用 nginx 或 Rust Web 服务器时也会发生同样的变化。在 M1、Linux docker 或 Linux EC2 机器上也是如此。是什么导致了从 10 微秒到 15 毫秒如此巨大的变化？谢谢

答案1

实际上，如果我每 5 毫秒执行一次 epoll/select，上限就会变得更低一些。

wrk -d2134 -t2 -c10 http://127.0.0.1:8089 -s script.lua               took  17s at  10:14:19
Running 36m test @ http://127.0.0.1:8089
  2 threads and 10 connections
^C  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    40.98us   18.24us   3.62ms   80.67%
    Req/Sec   109.03k     3.86k  115.50k    68.43%
  24368820 requests in 1.87m, 1.72GB read
Requests/sec: 216998.38
Transfer/sec:     15.73MB

JSON Output:
{
    "requests": 24368820,
    "duration_in_microseconds": 112299548.00,
    "bytes": 1852030320,
    "requests_per_sec": 216998.38,
    "bytes_transfer_per_sec": 16491876.89,
    "latency_distribution": [
        {
            "percentile": 0,
            "latency_in_microseconds": 10
        },
        {
            "percentile": 90,
            "latency_in_microseconds": 61
        },
        {
            "percentile": 99,
            "latency_in_microseconds": 88
        },
        {
            "percentile": 99.5,
            "latency_in_microseconds": 95
        },
        {
            "percentile": 99.9,
            "latency_in_microseconds": 117
        },
        {
            "percentile": 99.95,
            "latency_in_microseconds": 128
        },
        {
            "percentile": 99.99,
            "latency_in_microseconds": 230
        },
        {
            "percentile": 99.999,
            "latency_in_microseconds": 1803
        },
    ]
}

相关内容