多线程：如何合理地估算线程池大小？

我就是我 2023-06-14 11:56 85阅读 0赞

如何合理地估算线程池大小

这个问题虽然看起来很小，却并不那么容易回答。大家如果有更好的方法欢迎赐教，先来一个天真的估算方法：假设要求一个系统的TPS（Transaction Per Second或者Task Per Second）至少为20，然后假设每个Transaction由一个线程完成，继续假设平均每个线程处理一个Transaction的时间为4s。那么问题转化为：

如何设计线程池大小，使得可以在1s内处理完20个Transaction

计算过程很简单，每个线程的处理能力为0.25TPS，那么要达到20TPS，显然需要20/0.25=80个线程。

很显然这个估算方法很天真，因为它没有考虑到CPU数目。一般服务器的CPU核数为16或者32，如果有80个线程，那么肯定会带来太多不必要的线程上下文切换开销。

再来第二种简单的但不知是否可行的方法（N为CPU总核数）：

  • 如果是CPU密集型应用，则线程池大小设置为N+1
  • 如果是IO密集型应用，则线程池大小设置为2N+1

如果一台服务器上只部署这一个应用并且只有这一个线程池，那么这种估算或许合理，具体还需自行测试验证。

接下来在这个文档：服务器性能IO优化 中发现一个估算公式：

  1. 最佳线程数目 = （（线程等待时间+线程CPU时间）/线程CPU时间 ）* CPU数目

比如平均每个线程CPU运行时间为0.5s，而线程等待时间（非CPU运行时间，比如IO）为1.5s，CPU核心数为8，那么根据上面这个公式估算得到：((0.5+1.5)/0.5)*8=32。这个公式进一步转化为：

  1. 最佳线程数目 = （线程等待时间与线程CPU时间之比 + 1）* CPU数目

可以得出一个结论：

线程等待时间所占比例越高，需要越多线程。线程CPU时间所占比例越高，需要越少线程。

上一种估算方法也和这个结论相合。

一个系统最快的部分是CPU，所以决定一个系统吞吐量上限的是CPU。增强CPU处理能力，可以提高系统吞吐量上限。但根据短板效应，真实的系统吞吐量并不能单纯根据CPU来计算。那要提高系统吞吐量，就需要从“系统短板”（比如网络延迟、IO）着手：

  • 尽量提高短板操作的并行化比率，比如多线程下载技术
  • 增强短板能力，比如用NIO替代IO

第一条可以联系到Amdahl定律（阿姆达尔定律），这条定律定义了串行系统并行化后的加速比计算公式：

  1. 加速比=优化前系统耗时 / 优化后系统耗时

加速比越大，表明系统并行化的优化效果越好。Addahl定律（阿姆达尔定律）还给出了系统并行度、CPU数目和加速比的关系，加速比为Speedup，系统串行化比率（指串行执行代码所占比率）为F，CPU数目为N：

  1. Speedup <= 1 / (F + (1-F)/N)

当N足够大时，串行化比率F越小，加速比Speedup越大。

写到这里，我突然冒出一个问题。

是否使用线程池就一定比使用单线程高效呢？

答案是否定的，比如Redis就是单线程的，但它却非常高效，基本操作都能达到十万量级/s。从线程这个角度来看，部分原因在于：

  • 多线程带来线程上下文切换开销，单线程就没有这种开销
  • 锁

当然“Redis很快”更本质的原因在于：Redis基本都是内存操作，这种情况下单线程可以很高效地利用CPU。而多线程适用场景一般是：存在相当比例的IO和网络操作。

所以即使有上面的简单估算方法，也许看似合理，但实际上也未必合理，都需要结合系统真实情况（比如是IO密集型或者是CPU密集型或者是纯内存操作）和硬件环境（CPU、内存、硬盘读写速度、网络状况等）来不断尝试达到一个符合实际的合理估算值。

最后来一个“Dark Magic”估算方法（因为我暂时还没有搞懂它的原理），使用下面的类：

  1. package pool_size_calculate;
  3. import java.math.BigDecimal;
  4. import java.math.RoundingMode;
  5. import java.util.Timer;
  6. import java.util.TimerTask;
  7. import java.util.concurrent.BlockingQueue;
  9. /**
  10.  * A class that calculates the optimal thread pool boundaries. It takes the
  11.  * desired target utilization and the desired work queue memory consumption as
  12.  * input and retuns thread count and work queue capacity.
  13.  *
  14.  * @author Niklas Schlimm
  15.  *
  16.  */
  17. public abstract class PoolSizeCalculator {
  19.     /**
  20.      * The sample queue size to calculate the size of a single {@link Runnable}
  21.      * element.
  22.      */
  23.     private final int SAMPLE_QUEUE_SIZE = 1000;
  25.     /**
  26.      * Accuracy of test run. It must finish within 20ms of the testTime
  27.      * otherwise we retry the test. This could be configurable.
  28.      */
  29.     private final int EPSYLON = 20;
  31.     /**
  32.      * Control variable for the CPU time investigation.
  33.      */
  34.     private volatile boolean expired;
  36.     /**
  37.      * Time (millis) of the test run in the CPU time calculation.
  38.      */
  39.     private final long testtime = 3000;
  41.     /**
  42.      * Calculates the boundaries of a thread pool for a given {@link Runnable}.
  43.      *
  44.      * @param targetUtilization
  45.      *            the desired utilization of the CPUs (0 <= targetUtilization <=      *            1)      * @param targetQueueSizeBytes      *            the desired maximum work queue size of the thread pool (bytes)      */     protected void calculateBoundaries(BigDecimal targetUtilization,             BigDecimal targetQueueSizeBytes) {         calculateOptimalCapacity(targetQueueSizeBytes);         Runnable task = creatTask();         start(task);         start(task); // warm up phase         long cputime = getCurrentThreadCPUTime();         start(task); // test intervall         cputime = getCurrentThreadCPUTime() - cputime;         long waittime = (testtime * 1000000) - cputime;         calculateOptimalThreadCount(cputime, waittime, targetUtilization);     }     private void calculateOptimalCapacity(BigDecimal targetQueueSizeBytes) {         long mem = calculateMemoryUsage();         BigDecimal queueCapacity = targetQueueSizeBytes.divide(new BigDecimal(                 mem), RoundingMode.HALF_UP);         System.out.println("Target queue memory usage (bytes): "                 + targetQueueSizeBytes);         System.out.println("createTask() produced "                 + creatTask().getClass().getName() + " which took " + mem                 + " bytes in a queue");         System.out.println("Formula: " + targetQueueSizeBytes + " / " + mem);         System.out.println("* Recommended queue capacity (bytes): "                 + queueCapacity);     }     /**      * Brian Goetz' optimal thread count formula, see 'Java Concurrency in      * Practice' (chapter 8.2)      *       * @param cpu      *            cpu time consumed by considered task      * @param wait      *            wait time of considered task      * @param targetUtilization      *            target utilization of the system      */     private void calculateOptimalThreadCount(long cpu, long wait,             BigDecimal targetUtilization) {         BigDecimal waitTime = new BigDecimal(wait);         BigDecimal computeTime = new BigDecimal(cpu);         BigDecimal numberOfCPU = new BigDecimal(Runtime.getRuntime()                 .availableProcessors());         BigDecimal optimalthreadcount = numberOfCPU.multiply(targetUtilization)                 .multiply(                         new BigDecimal(1).add(waitTime.divide(computeTime,                                 RoundingMode.HALF_UP)));         System.out.println("Number of CPU: " + numberOfCPU);         System.out.println("Target utilization: " + targetUtilization);         System.out.println("Elapsed time (nanos): " + (testtime * 1000000));         System.out.println("Compute time (nanos): " + cpu);         System.out.println("Wait time (nanos): " + wait);         System.out.println("Formula: " + numberOfCPU + " * "                 + targetUtilization + " * (1 + " + waitTime + " / "                 + computeTime + ")");         System.out.println("* Optimal thread count: " + optimalthreadcount);     }     /**      * Runs the {@link Runnable} over a period defined in {@link #testtime}.      * Based on Heinz Kabbutz' ideas      * (http://www.javaspecialists.eu/archive/Issue124.html).      *       * @param task      *            the runnable under investigation      */     public void start(Runnable task) {         long start = 0;         int runs = 0;         do {             if (++runs > 5) {
  46.                 throw new IllegalStateException("Test not accurate");
  47.             }
  48.             expired = false;
  49.             start = System.currentTimeMillis();
  50.             Timer timer = new Timer();
  51.             timer.schedule(new TimerTask() {
  52.                 public void run() {
  53.                     expired = true;
  54.                 }
  55.             }, testtime);
  56.             while (!expired) {
  57.                 task.run();
  58.             }
  59.             start = System.currentTimeMillis() - start;
  60.             timer.cancel();
  61.         } while (Math.abs(start - testtime) > EPSYLON);
  62.         collectGarbage(3);
  63.     }
  65.     private void collectGarbage(int times) {
  66.         for (int i = 0; i < times; i++) {
  67.             System.gc();
  68.             try {
  69.                 Thread.sleep(10);
  70.             } catch (InterruptedException e) {
  71.                 Thread.currentThread().interrupt();
  72.                 break;
  73.             }
  74.         }
  75.     }
  77.     /**
  78.      * Calculates the memory usage of a single element in a work queue. Based on
  79.      * Heinz Kabbutz' ideas
  80.      * (http://www.javaspecialists.eu/archive/Issue029.html).
  81.      *
  82.      * @return memory usage of a single {@link Runnable} element in the thread
  83.      *         pools work queue
  84.      */
  85.     public long calculateMemoryUsage() {
  86.         BlockingQueue queue = createWorkQueue();
  87.         for (int i = 0; i < SAMPLE_QUEUE_SIZE; i++) {
  88.             queue.add(creatTask());
  89.         }
  90.         long mem0 = Runtime.getRuntime().totalMemory()
  91.                 - Runtime.getRuntime().freeMemory();
  92.         long mem1 = Runtime.getRuntime().totalMemory()
  93.                 - Runtime.getRuntime().freeMemory();
  94.         queue = null;
  95.         collectGarbage(15);
  96.         mem0 = Runtime.getRuntime().totalMemory()
  97.                 - Runtime.getRuntime().freeMemory();
  98.         queue = createWorkQueue();
  99.         for (int i = 0; i < SAMPLE_QUEUE_SIZE; i++) {
  100.             queue.add(creatTask());
  101.         }
  102.         collectGarbage(15);
  103.         mem1 = Runtime.getRuntime().totalMemory()
  104.                 - Runtime.getRuntime().freeMemory();
  105.         return (mem1 - mem0) / SAMPLE_QUEUE_SIZE;
  106.     }
  108.     /**
  109.      * Create your runnable task here.
  110.      *
  111.      * @return an instance of your runnable task under investigation
  112.      */
  113.     protected abstract Runnable creatTask();
  115.     /**
  116.      * Return an instance of the queue used in the thread pool.
  117.      *
  118.      * @return queue instance
  119.      */
  120.     protected abstract BlockingQueue createWorkQueue();
  122.     /**
  123.      * Calculate current cpu time. Various frameworks may be used here,
  124.      * depending on the operating system in use. (e.g.
  125.      * http://www.hyperic.com/products/sigar). The more accurate the CPU time
  126.      * measurement, the more accurate the results for thread count boundaries.
  127.      *
  128.      * @return current cpu time of current thread
  129.      */
  130.     protected abstract long getCurrentThreadCPUTime();
  132. }

然后自己继承这个抽象类并实现它的三个抽象方法，比如下面是我写的一个示例（任务是请求网络数据），其中我指定期望CPU利用率为1.0（即100%），任务队列总大小不超过100,000字节：

  1. package pool_size_calculate;
  3. import java.io.BufferedReader;
  4. import java.io.IOException;
  5. import java.io.InputStreamReader;
  6. import java.lang.management.ManagementFactory;
  7. import java.math.BigDecimal;
  8. import java.net.HttpURLConnection;
  9. import java.net.URL;
  10. import java.util.concurrent.BlockingQueue;
  11. import java.util.concurrent.LinkedBlockingQueue;
  13. public class SimplePoolSizeCaculatorImpl extends PoolSizeCalculator {
  15.     @Override
  16.     protected Runnable creatTask() {
  17.         return new AsyncIOTask();
  18.     }
  20.     @Override
  21.     protected BlockingQueue createWorkQueue() {
  22.         return new LinkedBlockingQueue(1000);
  23.     }
  25.     @Override
  26.     protected long getCurrentThreadCPUTime() {
  27.         return ManagementFactory.getThreadMXBean().getCurrentThreadCpuTime();
  28.     }
  30.     public static void main(String[] args) {
  31.         PoolSizeCalculator poolSizeCalculator = new SimplePoolSizeCaculatorImpl();
  32.         poolSizeCalculator.calculateBoundaries(new BigDecimal(1.0), new BigDecimal(100000));
  33.     }
  35. }
  37. /**
  38.  * 自定义的异步IO任务
  39.  * @author Will
  40.  *
  41.  */
  42. class AsyncIOTask implements Runnable {
  44.     @Override
  45.     public void run() {
  46.         HttpURLConnection connection = null;
  47.         BufferedReader reader = null;
  48.         try {
  49.             String getURL = "http://baidu.com";
  50.             URL getUrl = new URL(getURL);
  52.             connection = (HttpURLConnection) getUrl.openConnection();
  53.             connection.connect();
  54.             reader = new BufferedReader(new InputStreamReader(
  55.                     connection.getInputStream()));
  57.             String line;
  58.             while ((line = reader.readLine()) != null) {
  59.                 // empty loop
  60.             }
  61.         }
  63.         catch (IOException e) {
  65.         } finally {
  66.             if(reader != null) {
  67.                 try {
  68.                     reader.close();
  69.                 }
  70.                 catch(Exception e) {
  72.                 }
  73.             }
  74.             connection.disconnect();
  75.         }
  77.     }
  79. }

得到的输出如下：

  1. Target queue memory usage (bytes): 100000
  2. createTask() produced pool_size_calculate.AsyncIOTask which took 40 bytes in a queue
  3. Formula: 100000 / 40
  4. * Recommended queue capacity (bytes): 2500
  5. Number of CPU: 4
  6. Target utilization: 1
  7. Elapsed time (nanos): 3000000000
  8. Compute time (nanos): 47181000
  9. Wait time (nanos): 2952819000
  10. Formula: 4 * 1 * (1 + 2952819000 / 47181000)
  11. * Optimal thread count: 256

推荐的任务队列大小为2500，线程数为256，有点出乎意料之外。我可以如下构造一个线程池：

  1. ThreadPoolExecutor pool =
  2.  new ThreadPoolExecutor(256, 256, 0L, TimeUnit.MILLISECONDS, new LinkedBlockingQueue(2500));

发表评论

表情：
评论列表 （有 0 条评论，85人围观）

还没有评论，来说两句吧...

相关阅读

    相关 如何合理设置线程池大小

    接着上一篇探讨线程池留下的尾巴，如何合理的设置线程池大小。 要想合理的配置线程池的大小，首先得分析任务的特性，可以从以下几个角度分析： 1. 任务的性质：CPU密集型任

    女爷i女爷i/ 0 赞/ 347 阅读