Nothing Is Secret

Windows进程CPU、内存等资源限制

Windows自身没有提供类似Linux cgroup的能力来限制进程或进程组的资源占用，进程CPU/IO/内存/网络等资源的控制只能由自己实现。目前已有第三方的实现，主要是限制进程CPU的占用，如文档 < 21 Best Ways to Limit the CPU Usage of a Process > 所描述的BES，Process Tamer等软件。自Windows 8及Server 2012开始Windows系统有提供以job为单位的CPU占用及内存上限设置，之前的版本则只能以进程或线程为单位进行限制。

进程CPU占用限制方案

即时轮询系统所有进程(线程)的CPU占用，当发现所设定进程有超标时强制暂停进程所有线程的执行，然后在适当的时机再恢复执行。其中所涉及技术点：

进程CPU占用查询 GetProcessTimes

BOOL GetProcessTimes(
  [in]  HANDLE     hProcess,
  [out] LPFILETIME lpCreationTime,
  [out] LPFILETIME lpExitTime,
  [out] LPFILETIME lpKernelTime,
  [out] LPFILETIME lpUserTime
);

此函数可以获取进程从创建至当前的总运行时间及总的CPU时间，(KernelTime + UserTime) < 系统CPU数 * (当前时间 - CreationTime)

线程CPU占用查询 GetThreadTimes

BOOL GetThreadTimes(
  [in]  HANDLE     hThread,
  [out] LPFILETIME lpCreationTime,
  [out] LPFILETIME lpExitTime,
  [out] LPFILETIME lpKernelTime,
  [out] LPFILETIME lpUserTime
);

QueryThreadCycleTime可以提供更精准的CPU时间数据，单位为CPU时钟周期

BOOL QueryThreadCycleTime(
  [in]  HANDLE   ThreadHandle,
  [out] PULONG64 CycleTime
);

线程暂停及恢复

Windows平台没有提供暂停整个进程的支持函数，只能以线程为单位来操作，即SuspendThread及ResumeThread:

DWORD SuspendThread(
  [in] HANDLE hThread
);
DWORD ResumeThread(
  [in] HANDLE hThread
);

CPU亲和性设置: SetProcessAffinityMask

BOOL SetProcessAffinityMask(
  [in] HANDLE    hProcess,
  [in] DWORD_PTR dwProcessAffinityMask
);

此函数可以限定进程及其所有线程所能使用的CPU，故一定程序上亦限定了进程最大的系统CPU占用率。

DWORD_PTR SetThreadAffinityMask(
  [in] HANDLE    hThread,
  [in] DWORD_PTR dwThreadAffinityMask
);

此函数可单独限制特定线程的CPU亲和性。

进程优先级设置: SetPriorityClass

优先级解决的是优先运行及退让CPU的问题，本质上并不能限定CPU占用，只是优先级高于当前任务的忙碌的时候，当前进程会主动退让CPU 线程优先级设置：SetThreadPriority

BOOL SetThreadPriority(
  [in] HANDLE hThread,
  [in] int    nPriority
);

Job Objects

Windows系统提供了Job的概念用以管理多个进程，可以限制Job对象内所有进程及期线程的CPU核心占用、CPU占用及内存分配上限等，均通过SetInformationJobObject来实现，具体的CPU限制由JOBOBJECT_CPU_RATE_CONTROL_INFORMATION管理，内存限制则由JOBOBJECT_EXTENDED_LIMIT_INFORMATION来管理。

BOOL SetInformationJobObject(
  [in] HANDLE             hJob,
  [in] JOBOBJECTINFOCLASS JobObjectInformationClass,
  [in] LPVOID             lpJobObjectInformation,
  [in] DWORD              cbJobObjectInformationLength
);

需要注意的是CPU占用设置只有Windows 8及Server 2012之后的版本有效。

CPU Sets

此部分只限定了CPU Affinity属性

实验验证

可以直接利用开源项目go-winjob验证，验证系统Windows 8 x64，go-winjob git repo: https://github.com/kolesnikovae/go-winjob

验证程序

#include &lt;stdio.h&gt;
#include &lt;stdlib.h&gt;

void main(int argc, char *argv[])
{
        unsigned long total = 0, count = 0, i = 0;

        while (1) {
                if (malloc(1024)) {
                        total += 1024;
                        count++;
                }
                if (!(++i &amp; 4095))
                        printf(&quot;alloc: %u size: %u bytes\n&quot;, count, total);
    }
}

无限制

在无限制的情况下，此进程会占满一个CPU核心，commit内存总占用达2G CPUStress unlimited

单一进程

在设定CPU上限16%及内存16M上限之后，结果如下： CPUStress single process examples/job_object.go按如下修改：

var limits = []winjob.Limit{
        winjob.WithBreakawayOK(),
        winjob.WithKillOnJobClose(),
        winjob.WithActiveProcessLimit(3),
        winjob.WithProcessTimeLimit(10 * time.Second),
        winjob.WithCPUHardCapLimit(1600),        // 16%
        winjob.WithProcessMemoryLimit(16 &lt;&lt; 20), // 16MB
        winjob.WithWriteClipboardLimit(),
}

const defaultCommand = &quot;.\\CPUStress.exe&quot;

多进程(双进程)

将winjob.WithProcessMemoryLimit 改为 winjob.WithJobMemoryLimit，后者表示此job内所有进程要占用的总内存限制：

var limits = []winjob.Limit{
        winjob.WithBreakawayOK(),
        winjob.WithKillOnJobClose(),
        winjob.WithActiveProcessLimit(3),
        winjob.WithProcessTimeLimit(10 * time.Second),
        winjob.WithCPUHardCapLimit(1600),    // 16%
        winjob.WithJobMemoryLimit(16 &lt;&lt; 20), // 16MB
        winjob.WithWriteClipboardLimit(),
}

验证结果如下： CPUStress 2-processes CPUStress 2-processes

winjob example代码：

// +build windows

package main

import (
        &quot;encoding/json&quot;
        &quot;log&quot;
        &quot;os&quot;
        &quot;os/exec&quot;
        &quot;os/signal&quot;
        &quot;time&quot;

        &quot;golang.org/x/sys/windows&quot;

        &quot;github.com/kolesnikovae/go-winjob&quot;
)

var limits = []winjob.Limit{
        winjob.WithBreakawayOK(),
        winjob.WithKillOnJobClose(),
        winjob.WithActiveProcessLimit(3),
        winjob.WithProcessTimeLimit(10 * time.Second),
        winjob.WithCPUHardCapLimit(1600),    // 16%
        winjob.WithJobMemoryLimit(16 &lt;&lt; 20), // 16MB
        winjob.WithWriteClipboardLimit(),
}

const defaultCommand = &quot;.\\CPUStress.exe&quot;
const stressCommand  = &quot;.\\CPUStressX64.exe&quot;

func main() {
        job, err := winjob.Create(&quot;&quot;, limits...)
        if err != nil {
                log.Fatalf(&quot;Create: %v&quot;, err)
        }

        cmd := exec.Command(defaultCommand)
        cmd.Stderr = os.Stderr
        cmd.SysProcAttr = &amp;windows.SysProcAttr{
                CreationFlags: windows.CREATE_SUSPENDED,
        }
        if err := cmd.Start(); err != nil {
                log.Fatalf(&quot;Start: %v&quot;, err)
        }

        stress := exec.Command(stressCommand)
        stress.Stderr = os.Stderr
        stress.SysProcAttr = &amp;windows.SysProcAttr{
                CreationFlags: windows.CREATE_SUSPENDED,
        }
        if err := stress.Start(); err != nil {
                log.Fatalf(&quot;Start: %v&quot;, err)
        }

        s := make(chan os.Signal, 1)
        signal.Notify(s, os.Interrupt)

        c := make(chan winjob.Notification)
        subscription, err := winjob.Notify(c, job)
        if err != nil {
                log.Fatalf(&quot;Notify: %v&quot;, err)
        }

        done := make(chan struct{})
        go func() {
                defer close(done)
                ticker := time.NewTicker(time.Second * 5)
                defer ticker.Stop()
                var counters winjob.Counters
                for {
                        select {
                        case &lt;-s:
                                log.Println(&quot;Closing job object&quot;)
                                if err := job.Close(); err != nil {
                                        log.Fatal(err)
                                }
                                log.Println(&quot;Closing subscription&quot;)
                                if err := subscription.Close(); err != nil {
                                        log.Fatal(err)
                                }
                                return

                        case n, ok := &lt;-c:
                                if ok {
                                        log.Printf(&quot;Notification: %#v\n&quot;, n)
                                } else if err := subscription.Err(); err != nil {
                                        log.Fatalf(&quot;Subscription: %v&quot;, err)
                                }

                        case &lt;-ticker.C:
                                if err := job.QueryCounters(&amp;counters); err != nil {
                                        log.Fatalf(&quot;QueryCounters: %v&quot;, err)
                                }
                                b, err := json.MarshalIndent(counters, &quot;&quot;, &quot;\t&quot;)
                                if err != nil {
                                        log.Fatal(err)
                                }
                                log.Printf(&quot;Counters: \n%s\n&quot;, b)
                        }
                }
        }()

        if err := job.Assign(cmd.Process); err != nil {
                log.Fatalf(&quot;Assign: %v&quot;, err)
        }
        if err := winjob.Resume(cmd); err != nil {
                log.Fatalf(&quot;Resume: %v&quot;, err)
        }

        if err := job.Assign(stress.Process); err != nil {
                log.Fatalf(&quot;Assign: %v&quot;, err)
        }
        if err := winjob.Resume(stress); err != nil {
                log.Fatalf(&quot;Resume: %v&quot;, err)
        }

        if err := cmd.Wait(); err != nil {
                log.Fatalf(&quot;Wait: %v&quot;, err)
        }
        if err := stress.Wait(); err != nil {
                log.Fatalf(&quot;Wait: %v&quot;, err)
        }

        // Wait for a signal.
        &lt;-done
}

参考链接

  1. 21 Best Ways to Limit the CPU Usage of a Process
  2. MSDN: Windows Process and Thread Functions
  3. MSDN: CPU Sets
  4. GetThreadTimes

71,948 条评论

  1. I’ve been experimenting with Rapid Indexer VIP recently to speed up my workflows. It’s hitting about an 89% success rate, which is decent, but it’s no silver bullet. I’m still seeing plenty of pages hit the ‘crawled – currently not indexed’ status in GSC how to use url inspection api

  12. UK88 được xây dựng theo hướng một nền tảng nội dung có cấu trúc rõ ràng và dễ tiếp cận, giúp người truy cập nhanh chóng xác định khu vực mình quan tâm bbc

  14. SIN88 được xây dựng theo hướng một nền tảng nội dung có cấu trúc rõ ràng và dễ tiếp cận, giúp người truy cập nhanh chóng xác định khu vực mình quan tâm bbc

  15. VIN88 được xây dựng theo hướng một nền tảng nội dung có cấu trúc rõ ràng và dễ tiếp cận, giúp người truy cập nhanh chóng xác định khu vực mình quan tâm bbc

  16. XOILAC được xây dựng theo hướng một nền tảng nội dung có cấu trúc rõ ràng và dễ tiếp cận, giúp người truy cập nhanh chóng xác định khu vực mình quan tâm bbc

  32. It is true that breeds like French Bulldogs are pricey upfront, but the recurring medical costs really add up. My biggest piece of advice for new owners is to sort out pet insurance as early as possible Visit this link

  33. I’ve been testing various indexing methods lately, and while hitting an 89% success rate sounds great on paper, it’s not always a silver bullet. Even with these tools, I still run into the classic ‘crawled – currently not indexed’ status in GSC constantly indexceptional refund policy

  42. This $9 billion projection really hits home. We see massive potential in repurposing live session footage into snackable micro-content after the event ends. It keeps the audience engaged long after they log off Great site

  45. It is true that breeds like the French Bulldog can be quite pricey to maintain due to their specific health needs. I always recommend getting lifetime insurance cover as soon as you bring them home because those vet bills add up quickly Additional reading

  49. I’ve been looking at ManyPets for my two-year-old spaniel since our renewal is coming up in March. That 24/7 video vet feature sounds really handy for peace of mind, especially during those late-night panics Additional hints

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注