如何合理地估算线程池大小?(转载)
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:
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:
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:
第一条可以联系到Amdahl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:
加速比越大,表明系统并行化的优化效果越好。Addahl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:
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:
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 threadpool; 2 3 import java.math.BigDecimal; 4 java.math.RoundingMode; 5 java.util.Timer; 6 java.util.TimerTask; 7 java.util.concurrent.BlockingQueue; 8 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 public abstract class PoolSizeCalculator { 17 18 19 * The sample queue size to calculate the size of a single {@link Runnable} 20 * element. 21 22 private final int SAMPLE_QUEUE_SIZE = 1000; 23 24 25 * Accuracy of test run. It must finish within 20ms of the testTime 26 * otherwise we retry the test. This could be configurable. 27 28 int EPSYLON = 20 29 30 31 * Control variable for the CPU time investigation. 32 33 volatile boolean expired; 34 35 36 * Time (millis) of the test run in the CPU time calculation. 37 38 long testtime = 3000 39 40 41 * Calculates the boundaries of a thread pool for a given { Runnable}. 42 * 43 * @param targetUtilization the desired utilization of the CPUs (0 <= targetUtilization <= * 1) * targetQueueSizeBytes * the desired maximum work queue size of the thread pool (bytes) 44 45 protected void calculateBoundaries(BigDecimal targetUtilization,BigDecimal targetQueueSizeBytes) { 46 calculateOptimalCapacity(targetQueueSizeBytes); 47 Runnable task = creatTask(); 48 start(task); 49 start(task); // warm up phase 50 long cputime = getCurrentThreadCPUTime(); 51 start(task); test intervall 52 cputime = getCurrentThreadCPUTime() - cputime; 53 long waittime = (testtime * 1000000) - 54 calculateOptimalThreadCount(cputime,waittime,targetUtilization); 55 } 56 57 calculateOptimalCapacity(BigDecimal targetQueueSizeBytes) { 58 long mem = calculateMemoryUsage(); 59 BigDecimal queueCapacity = targetQueueSizeBytes.divide(new BigDecimal(mem), 60 RoundingMode.HALF_UP); 61 System.out.println("Target queue memory usage (bytes): " 62 + targetQueueSizeBytes); 63 System.out.println("createTask() produced " + creatTask().getClass().getName() + " which took " + mem + " bytes in a queue"); 64 System.out.println("Formula: " + targetQueueSizeBytes + " / " + mem); 65 System.out.println("* Recommended queue capacity (bytes): " + queueCapacity); 66 67 68 69 * Brian Goetz' optimal thread count formula,see 'Java Concurrency in 70 * * Practice' (chapter 8.2) * 71 * * cpu 72 * * cpu time consumed by considered task 73 wait 74 * * wait time of considered task 75 targetUtilization 76 * * target utilization of the system 77 78 void calculateOptimalThreadCount(long cpu,long wait,1)"> 79 BigDecimal targetUtilization) { 80 BigDecimal waitTime = BigDecimal(wait); 81 BigDecimal computeTime = BigDecimal(cpu); 82 BigDecimal numberOfCPU = BigDecimal(Runtime.getRuntime() 83 .availableProcessors()); 84 BigDecimal optimalthreadcount = numberOfCPU.multiply(targetUtilization) 85 .multiply(new BigDecimal(1).add(waitTime.divide(computeTime,1)"> 86 RoundingMode.HALF_UP))); 87 System.out.println("Number of CPU: " + numberOfCPU); 88 System.out.println("Target utilization: " + targetUtilization); 89 System.out.println("Elapsed time (nanos): " + (testtime * 1000000)); 90 System.out.println("Compute time (nanos): " + cpu); 91 System.out.println("Wait time (nanos): " + wait); 92 System.out.println("Formula: " + numberOfCPU + " * " 93 + targetUtilization + " * (1 + " + waitTime + " / " 94 + computeTime + ")" 95 System.out.println("* Optimal thread count: " + optimalthreadcount); 96 97 98 99 * * Runs the { Runnable} over a period defined in { #testtime}. 100 * * Based on Heinz Kabbutz' ideas 101 * * (http://www.javaspecialists.eu/archive/Issue124.html). 102 * * 103 task 104 * * the runnable under investigation 105 106 start(Runnable task) { 107 long start = 0108 int runs = 0109 do { 110 if (++runs > 5) { 111 throw new IllegalStateException("Test not accurate"112 } 113 expired = false114 start = System.currentTimeMillis(); 115 Timer timer = Timer(); 116 timer.schedule( TimerTask() { 117 run() { 118 expired = true119 } 120 },testtime); 121 while (!expired) { 122 task.run(); 123 124 start = System.currentTimeMillis() - start; 125 timer.cancel(); 126 } while (Math.abs(start - testtime) > EPSYLON); 127 collectGarbage(3128 129 130 void collectGarbage(int times) { 131 for (int i = 0; i < times; i++132 System.gc(); 133 try134 Thread.sleep(10135 } catch (InterruptedException e) { 136 Thread.currentThread().interrupt(); 137 break138 139 } 140 141 142 143 * Calculates the memory usage of a single element in a work queue. Based on 144 * Heinz Kabbutz' ideas 145 * (http://www.javaspecialists.eu/archive/Issue029.html146 147 @return memory usage of a single { Runnable} element in the thread 148 * pools work queue 149 150 calculateMemoryUsage() { 151 BlockingQueue queue = createWorkQueue(); 152 int i = 0; i < SAMPLE_QUEUE_SIZE; i++153 queue.add(creatTask()); 154 155 156 long mem0 = Runtime.getRuntime().totalMemory() - Runtime.getRuntime().freeMemory(); 157 long mem1 = Runtime.getRuntime().totalMemory() -158 159 queue = null160 161 collectGarbage(15162 163 mem0 = Runtime.getRuntime().totalMemory() -164 queue =165 166 167 168 169 170 collectGarbage(15171 172 mem1 = Runtime.getRuntime().totalMemory() -173 174 return (mem1 - mem0) / SAMPLE_QUEUE_SIZE; 175 176 177 178 * Create your runnable task here. 179 180 an instance of your runnable task under investigation 181 182 abstract Runnable creatTask(); 183 184 185 * Return an instance of the queue used in the thread pool. 186 187 queue instance 188 189 BlockingQueue createWorkQueue(); 190 191 192 * Calculate current cpu time. Various frameworks may be used here,1)">193 * depending on the operating system in use. (e.g. 194 * http://www.hyperic.com/products/sigar). The more accurate the CPU time 195 * measurement,the more accurate the results for thread count boundaries. 196 197 current cpu time of current thread 198 199 200 201 } 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 2 3 java.io.BufferedReader; 4 java.io.IOException; 5 java.io.InputStreamReader; 6 java.lang.management.ManagementFactory; 7 8 java.net.HttpURLConnection; 9 java.net.URL; 10 11 java.util.concurrent.LinkedBlockingQueue; 12 13 class SimplePoolSizeCaculatorImpl extends14 15 @Override 16 protected Runnable creatTask() { 17 return AsyncIOTask(); 18 19 20 21 BlockingQueue createWorkQueue() { 22 new LinkedBlockingQueue(100023 24 25 26 getCurrentThreadCPUTime() { 27 return ManagementFactory.getThreadMXBean().getCurrentThreadCpuTime(); 28 29 30 static main(String[] args) { 31 PoolSizeCalculator poolSizeCalculator = SimplePoolSizeCaculatorImpl(); 32 poolSizeCalculator.calculateBoundaries(new BigDecimal(1.0),1)">new BigDecimal(10000033 34 35 } 36 37 38 * 自定义的异步IO任务 39 Will 40 41 42 class AsyncIOTask implements Runnable { 43 44 45 HttpURLConnection connection = 46 BufferedReader reader = 47 48 String getURL = "http://baidu.com"49 URL getUrl = URL(getURL); 50 51 connection = (HttpURLConnection) getUrl.openConnection(); 52 connection.connect(); 53 reader = new BufferedReader( InputStreamReader( 54 connection.getInputStream())); 55 56 String line; 57 while ((line = reader.readLine()) != 58 empty loop 59 60 61 62 (IOException e) { 63 64 } finally65 if(reader != 66 67 reader.close(); 68 69 (Exception e) { 70 71 72 73 connection.disconnect(); 74 75 76 77 78 } 得到如下输出: Target queue memory usage (bytes): 100000 createTask() produced threadpool.AsyncIOTask which took 40 bytes in a queue Formula: 100000 / 40 * Recommended queue capacity (bytes): 2500 Number of CPU: 8 Target utilization: 1 Elapsed time (nanos): 3000000000 Compute time (nanos): 280801800 Wait time (nanos): 2719198200 Formula: 8 * 1 * (1 + 2719198200 / 280801800) * Optimal thread count: 88 推荐的任务队列大小为2500,线程数为88。依次为依据,我们就可以构造这样一个线程池: ThreadPoolExecutor pool = new ThreadPoolExecutor(88,88,0L,TimeUnit.MILLISECONDS,1)">new LinkedBlockingQueue<Runnable>(2500));
可以将这个文件打包成可执行的jar文件,这样就可以拷贝到测试/正式环境上执行。 <project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" 2 xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd"> 3 modelVersion>4.0.0</ 4 5 groupId>threadpool 6 artifactId>dark-magic 7 version>1.0-SNAPSHOT 8 packaging>jar 9 10 name>dark_magic11 url>http://maven.apache.org13 properties14 project.build.sourceEncoding>UTF-815 16 17 dependencies18 19 20 buildfinalName23 24 plugins25 plugin26 >maven-assembly-plugin27 configuration28 appendAssemblyId>false29 descriptorRefs30 descriptorRef>jar-with-dependencies31 32 archive33 manifest34 <!-- 此处指定main方法入口的class --> 35 mainClass>threadpool.SimplePoolSizeCaculatorImpl36 37 38 39 executions40 execution41 id>make-assembly42 phase>package43 goals44 goal>assembly45 46 47 48 49 50 51 project> ? 转载: http://ifeve.com/how-to-calculate-threadpool-size/ http://www.importnew.com/17384.html https://www.cnblogs.com/cherish010/p/8334952.html ? (编辑:李大同) 【声明】本站内容均来自网络,其相关言论仅代表作者个人观点,不代表本站立场。若无意侵犯到您的权利,请及时与联系站长删除相关内容! |