测试HashTable、Collections.synchronizedMap和ConcurrentHashMap的性能

    对于map的并发操作有HashTable、Collections.synchronizedMap和ConcurrentHashMap三种，到底性能如何呢？

测试代码：

package com.yangyang;

import java.util.Collections;
import java.util.HashMap;
import java.util.Hashtable;
import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;

public class T {
    /**用于测试的线程数量**/
    public static final int threads = 100;
    /**每个线程往map中塞的数量**/
    public static final int NUMBER =100;
      
    public static void main(String[] args) throws Exception{
        Map<String, Integer> syncHashMap=Collections.synchronizedMap(new HashMap<String, Integer>());
        Map<String, Integer> concurrentHashMap=new ConcurrentHashMap<String, Integer>();
        Hashtable<String, Integer> hashtable=new Hashtable<String, Integer>();
        
        long totalA = 0;
        long totalB = 0;
        long totalC = 0;
        //循环10此，累计时间，便于观察
        for (int i = 0; i <= 10; i++) {
        // System.out.println("第"+i+"次测试put方法");
          totalA += testPut(syncHashMap);
          totalB += testPut(concurrentHashMap);
          totalC += testPut(hashtable);
        }
        System.out.println("Put time HashMapSync=" + totalA + "ms.");
        System.out.println("Put time ConcurrentHashMap=" + totalB + "ms.");
        System.out.println("Put time Hashtable=" + totalC + "ms.");
        
        totalA = 0;
        totalB = 0;
        totalC = 0;
        for (int i = 0; i <= 10; i++) {
          totalA += testGet(syncHashMap);
          totalB += testGet(concurrentHashMap);
          totalC += testGet(hashtable);
        }
        System.out.println("Get time HashMapSync=" + totalA + "ms.");
        System.out.println("Get time ConcurrentHashMap=" + totalB + "ms.");
        System.out.println("Get time Hashtable=" + totalC + "ms.");
        
    }

    private static long testPut(Map<String, Integer> map) throws Exception{
        long start = System.currentTimeMillis();
        
        //同时开threads个线程
        for (int i = 0; i < threads; i++) {
            new MapPutThread(map).start();
        }
        while (MapPutThread.counter > 0) {
          Thread.sleep(1);
        }
        return System.currentTimeMillis() - start;
    }
    
     public static long testGet(Map<String, Integer> map) throws Exception {
        long start = System.currentTimeMillis();
        for (int i = 0; i < threads; i++) {
          new MapGetThread(map).start();
        }
        while (MapGetThread.counter > 0) {
          Thread.sleep(1);
        }
        return System.currentTimeMillis() - start;
     }
}
/**
 * put线程类
 * @author shunyang
 * @date 2015年3月6日 下午4:24:42
 */
class MapPutThread extends Thread{

      static int counter = 0;//计数器
      static Object lock = new Object();//用于同步的对象锁
      private Map<String, Integer> map;
      private String key = this.getId() + "";
      
      MapPutThread(Map<String, Integer> map) {
        synchronized (lock) {
          counter++;//每调用一次构建函数，计数器加1
     //     System.out.println("线程key为："+key+"的构造函数运行，当前counter为："+counter);
        }
        this.map = map;
      }
      
      
      @Override
      public void run() {
        for (int i = 1; i <= T.NUMBER; i++) {
          map.put(key, i);
        //  System.out.println("线程key为："+key+"的第"+i+"个run方法运行，设置的i为：:"+i);
        }
        synchronized (lock) {
          counter--;//每当往map中put一个值后，计算器减1
       //   System.out.println("线程key为："+key+"的run()运行，当前counter为："+counter);
        }
      }
}
/**
 * get线程类    
 * @author shunyang
 * @date 2015年3月6日 下午4:24:52
 */
class MapGetThread extends Thread {
    
  static int counter = 0;
  static final Object lock = new Object();
  private Map<String, Integer> map;
  private String key = this.getId() + "";
  
  MapGetThread(Map<String, Integer> map) {
    synchronized (lock) {
      counter++;
    }
    this.map = map;
  }
  
  @Override
  public void run() {
    for (int i = 1; i <= T.NUMBER; i++) {
      map.get(key);
    }
    synchronized (lock) {
      counter--;
    }
  }
}

当每次启动100个线程，每个线程往map中塞100个数据的时候，结果：

当每次启动1000个线程，每个线程往map中塞1000个数据的时候，结果：

当每次启动10000个线程，每个线程往map中塞10000个数据的时候，结果：

结论：当线程越多时，

ConcurrentHashMap的性能比同步的HashMap快一倍左右
同步的HashMap和Hashtable的性能相当