线程等待——CountDownLatch使用

告警性能优化过程中，遇到如下问题：1、 在数据库计算几十万个实体的KPI值的方差；2、 计算结果进行表格化处理。

这里KPI包含多个Counter的数据库函数运算（比如Decode，AVG等函数），方差也是数据库函数运算，性能比较差。

步骤1中每个实体独立计算方差，步骤2需要方差结果协同处理，所以很自然的联想到步骤1分实体多线程处理，步骤2等待步骤1所有线程完成后才开始处理。这里我们使用CountDownLatch进行线程等待，示例代码如下：

package com.coshaho.threadpool;

import java.util.concurrent.CountDownLatch;

/**
 * CountDownLatch学习
 * @author coshaho
 */
public class MyCountDownLatch
{
	public static void main(String[] args) throws InterruptedException
	{
		// 定义线程等待变量CountDownLatch，此处定义等待3个线程执行完成
		CountDownLatch latch = new CountDownLatch(3);
		// 定义3个线程，并传入线程等待变量
		new Thread(new MyCountDownLatch().new MyWork("Thread1",latch)).start();
		new Thread(new MyCountDownLatch().new MyWork("Thread2",latch)).start();
		new Thread(new MyCountDownLatch().new MyWork("Thread3",latch)).start();
		// 等待3个线程执行完成
		latch.await();
		System.out.println("All works are done.");
	}

	/**
	 * 线程任务
	 * @author coshaho
	 */
	private class MyWork implements Runnable
	{
		private String workName;
		private CountDownLatch latch;
		public MyWork(String workName, CountDownLatch latch)
		{
			this.workName = workName;
			this.latch = latch;
		}

		@Override
		public void run()
		{
			try
			{
				System.out.println("Thread " + workName + " is running.");
				Thread.sleep(1000);
				System.out.println("Thread " + workName + " is stop.");
			}
			catch (InterruptedException e)
			{
				e.printStackTrace();
			}
			finally
			{
				// 线程执行完成，线程等待变量减少
				latch.countDown();
			}
		}
	}
}

运行结果如下：

Thread Thread1 is running.
Thread Thread3 is running.
Thread Thread2 is running.
Thread Thread1 is stop.
Thread Thread3 is stop.
Thread Thread2 is stop.
All works are done.