如何实现线程的HashMap以及性能对比

对于线程是否是安全的问题，一般是不会用到的，因为在开发中通用的是单线程倒是不用担心线程安全问题。

但是对于要进行高并发的开发，线程安全就必不可少了，这也是面试的拦路虎，在此总结一下前人的经验，进行线程安全的总结，如有不足的地方请指出交流。

借鉴别人的测试代码并完善一下:

import java.util.Collections;
import java.util.HashMap;
import java.util.Hashtable;
import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentMap;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicLong;

public class CrunchifyConcurrentHashMapVsSynchronizedMap {

	public final static int THREAD_POOL_SIZE = 10;
	public final static int THREAD_LOOP_TIME = 100000;
	public final static String key = "wordcounts" ;

	
	public static Map<String, AtomicLong> crunchifyHashTableObject = null;

	public static Map<String, AtomicLong> crunchifySynchronizedMapObject = null;

	public static Map<String, AtomicLong> crunchifyConcurrentHashMapObject = null;
	public static Map<String, AtomicLong> crunchifyHashMapObject = null;

	public static void main(String[] args) throws

	InterruptedException {
		// 安全性测试
		mapSecurictyTest();
		//效率测试
		mapEfficienceTest();
	}

	private static void mapSecurictyTest() throws InterruptedException {
		// Test with HashMap Object
		crunchifyHashMapObject = new HashMap<>();
		threadSecurityTest(crunchifyHashMapObject);
				
		// Test with Hashtable Object
		crunchifyHashTableObject = new Hashtable<>();
		threadSecurityTest(crunchifyHashTableObject);

		// Test with synchronizedMap Object
		crunchifySynchronizedMapObject = Collections.synchronizedMap(new HashMap<String, AtomicLong>());
		threadSecurityTest(crunchifySynchronizedMapObject);

		// Test with ConcurrentHashMap Object
		crunchifyConcurrentHashMapObject = new ConcurrentHashMap<>();
		threadSecurityTest(crunchifyConcurrentHashMapObject);
		
	}

	private static void mapEfficienceTest() throws InterruptedException {
		
		// 效率最高 38ms
		Map<String, Integer> crunchifyHashMapObject = null;
		
		//	For class java.util.Hashtable the average time is 82 ms
		Map<String, Integer> crunchifyHashTableObject = null;
		//	For class java.util.Collections$SynchronizedMap the average time is 84 ms
		Map<String, Integer> crunchifySynchronizedMapObject = null;
		//	For class java.util.concurrent.ConcurrentHashMap the average time is 58 ms
		Map<String, Integer> crunchifyConcurrentHashMapObject = null;

		// Test with HashMap Object
		crunchifyHashMapObject = new HashMap<>();
		crunchifyPerformTest(crunchifyHashMapObject);

		// Test with Hashtable Object
		crunchifyHashTableObject = new Hashtable<>();
		crunchifyPerformTest(crunchifyHashTableObject);

		// Test with synchronizedMap Object
		crunchifySynchronizedMapObject = Collections.synchronizedMap(new HashMap<String, Integer>());
		crunchifyPerformTest(crunchifySynchronizedMapObject);

		// Test with ConcurrentHashMap Object
		crunchifyConcurrentHashMapObject = new ConcurrentHashMap<>();
		crunchifyPerformTest(crunchifyConcurrentHashMapObject);
	}
	
	public static void threadSecurityTest( final Map<String, AtomicLong> map) throws InterruptedException {

		System.out.println("Test started for: " + map.getClass());
		long averageTime = 0;
		for (int i = 0; i < 5; i++) {
			long startTime = System.nanoTime();
			ExecutorService crunchifyExServer = Executors.newFixedThreadPool(THREAD_POOL_SIZE);
			for (int j = 0; j < THREAD_POOL_SIZE; j++) {
				crunchifyExServer.execute(new Runnable() {
					@Override
					public void run() {

						for (int i = 0; i < THREAD_LOOP_TIME; i++) {
						    AtomicLong number = map.get(key);
						    if (number == null) {
						        AtomicLong newNumber = new AtomicLong(0);
						        number = map.putIfAbsent(key, newNumber);
						        if (number == null) {
						            number = newNumber;
						        }
						    }
					        long incrementValue = number.incrementAndGet();
						        
							if ((incrementValue )%500000==00) {
								System.out.println("value is :" +(incrementValue ));
							}

						}
					}
				});
			}

			// Make sure executor stops
			crunchifyExServer.shutdown();
			// Blocks until all tasks have completed execution after a shutdown  request
			crunchifyExServer.awaitTermination(Long.MAX_VALUE, TimeUnit.DAYS);

			long entTime = System.nanoTime();
			long totalTime = (entTime - startTime) / 1000000L;
			averageTime += totalTime;

			System.out.println( (i+1 )+"-"+THREAD_POOL_SIZE*THREAD_LOOP_TIME +" entried added/retrieved in " + totalTime + " ms");
		}
		System.out.println("For " + map.getClass() + " the average time is " + averageTime / 5 + " ms\n");
	}

	public static void crunchifyPerformTest( final Map<String, Integer> crunchifyThreads) throws InterruptedException {

		System.out.println("Test started for: " + crunchifyThreads.getClass());

		long averageTime = 0;

		for (int i = 0; i < 5; i++) {

			long startTime = System.nanoTime();
			ExecutorService crunchifyExServer = Executors.newFixedThreadPool(THREAD_POOL_SIZE);
			for (int j = 0; j < THREAD_POOL_SIZE; j++) {

				crunchifyExServer.execute(new Runnable() {

					@Override
					public void run() {

						for (int i = 0; i < THREAD_LOOP_TIME; i++) {

 
							// Retrieve value. We are not using it anywhere
							Integer crunchifyValue = crunchifyThreads.get(key);
							
							// Put value
							crunchifyThreads.put( key, new Integer(1));
						}
					}
				});
			}
			// Make sure executor stops
			crunchifyExServer.shutdown();
			// Blocks until all tasks have completed execution after a shutdown  request
			crunchifyExServer.awaitTermination(Long.MAX_VALUE, TimeUnit.DAYS);

			long entTime = System.nanoTime();
			long totalTime = (entTime - startTime) / 1000000L;
			averageTime += totalTime;

			System.out.println("2500K entried added/retrieved in " + totalTime + " ms");
		}
		System.out.println("For " + crunchifyThreads.getClass() + " the average time is " + averageTime / 5 + " ms\n");
	}
	
	 
	
}

测试结果：

由打印结果发现 HashMap居然能够累加到500w.难道HashMap也能线程安全吗？

================================== map securicty test start ==================================
Test started for: class java.util.HashMap
value is :500000
value is :1000000
1-1000000 entried added/retrieved in 57 ms
value is :1500000
value is :2000000
2-1000000 entried added/retrieved in 51 ms
value is :2500000
value is :3000000
3-1000000 entried added/retrieved in 38 ms
value is :3500000
value is :4000000
4-1000000 entried added/retrieved in 35 ms
value is :4500000
value is :5000000
5-1000000 entried added/retrieved in 35 ms
For class java.util.HashMap the average time is 43 ms


Test started for: class java.util.Hashtable
value is :500000
value is :1000000
1-1000000 entried added/retrieved in 97 ms
value is :1500000
value is :2000000
2-1000000 entried added/retrieved in 66 ms
value is :2500000
value is :3000000
3-1000000 entried added/retrieved in 67 ms
value is :3500000
value is :4000000
4-1000000 entried added/retrieved in 65 ms
value is :4500000
value is :5000000
5-1000000 entried added/retrieved in 61 ms
For class java.util.Hashtable the average time is 71 ms


Test started for: class java.util.Collections$SynchronizedMap
value is :500000
value is :1000000
1-1000000 entried added/retrieved in 111 ms
value is :1500000
value is :2000000
2-1000000 entried added/retrieved in 90 ms
value is :2500000
value is :3000000
3-1000000 entried added/retrieved in 92 ms
value is :3500000
value is :4000000
4-1000000 entried added/retrieved in 94 ms
value is :4500000
value is :5000000
5-1000000 entried added/retrieved in 98 ms
For class java.util.Collections$SynchronizedMap the average time is 97 ms


Test started for: class java.util.concurrent.ConcurrentHashMap
value is :500000
value is :1000000
1-1000000 entried added/retrieved in 40 ms
value is :1500000
value is :2000000
2-1000000 entried added/retrieved in 35 ms
value is :2500000
value is :3000000
3-1000000 entried added/retrieved in 32 ms
value is :3500000
value is :4000000
4-1000000 entried added/retrieved in 34 ms
value is :4500000
value is :5000000
5-1000000 entried added/retrieved in 33 ms
For class java.util.concurrent.ConcurrentHashMap the average time is 34 ms


================================== map securicty test end ==================================

带着疑问再次进行单独测试：

Test started for: class java.util.HashMap
value is :500000
1-1000000 entried added/retrieved in 51 ms
value is :1000000
value is :1500000
2-1000000 entried added/retrieved in 40 ms
value is :2000000
value is :2500000
3-1000000 entried added/retrieved in 31 ms
value is :3000000
value is :3500000
4-1000000 entried added/retrieved in 30 ms
value is :4000000
value is :4500000
5-1000000 entried added/retrieved in 30 ms
For class java.util.HashMap the average time is 36 ms

发现这次就没有能够达到线程安全。

所以线程安全的策略有三种，而且要和atomic结合操作才能实现。

效率测试如下：

HashMap的效率最高，如果不考虑线程的问题（单线程使用的时候），HashMap是首选。

其次是ConcurrentHashMap，Collections$SynchronizedMap,Hashtable 。

一般情况下ConcurrentHashMap效率能够达到Collections$SynchronizedMap或者Hashtable 的两倍，
================================== map Efficience test start ==================================
Test started for: class java.util.HashMap
2500K entried added/retrieved in 58 ms
2500K entried added/retrieved in 49 ms
2500K entried added/retrieved in 26 ms
2500K entried added/retrieved in 27 ms
2500K entried added/retrieved in 28 ms
For class java.util.HashMap the average time is 37 ms


Test started for: class java.util.Hashtable
2500K entried added/retrieved in 113 ms
2500K entried added/retrieved in 76 ms
2500K entried added/retrieved in 72 ms
2500K entried added/retrieved in 75 ms
2500K entried added/retrieved in 74 ms
For class java.util.Hashtable the average time is 82 ms


Test started for: class java.util.Collections$SynchronizedMap
2500K entried added/retrieved in 100 ms
2500K entried added/retrieved in 87 ms
2500K entried added/retrieved in 88 ms
2500K entried added/retrieved in 84 ms
2500K entried added/retrieved in 87 ms
For class java.util.Collections$SynchronizedMap the average time is 89 ms


Test started for: class java.util.concurrent.ConcurrentHashMap
2500K entried added/retrieved in 63 ms
2500K entried added/retrieved in 58 ms
2500K entried added/retrieved in 58 ms
2500K entried added/retrieved in 57 ms
2500K entried added/retrieved in 60 ms
For class java.util.concurrent.ConcurrentHashMap the average time is 59 ms


================================== map Efficience test end ==================================

参考： http://blog.youkuaiyun.com/cnhk1225/article/details/52413201