线程池ExecutorService 及CountDownLatch 并发执行查询接口

public static Map<String ,List<Map<String, Object>>> getMaps(List<String> sqlKeys) throws InterruptedException, ExecutionException{
	    if(CollectionUtils.isEmpty(sqlKeys)) {
		return null;
	    }
	    Map<String ,List<Map<String, Object>>> map = new HashMap<String ,List<Map<String, Object>>>();
	    ExecutorService executorService = Executors.newFixedThreadPool(2);
	    final CountDownLatch countDownLatch = new CountDownLatch(2);
	    Future<List<Map<String, Object>>> outputDatas = executorService.submit(new Callable<List<Map<String, Object>>>() {
	    @Override
	    public List<Map<String, Object>> call() throws Exception {
	    		try {
	    			return new HanaJdbc().querySQLForList(PropertyPlaceholderUtil.getProperty("COMPANY.CODE"), null, false);
	    		}finally {
	    			countDownLatch.countDown();
				}
	    	}
		});
	    
	    Future<List<Map<String, Object>>> outputDatas2 = executorService.submit(new Callable<List<Map<String, Object>>>() {
	    	@Override
	    	public List<Map<String, Object>> call() throws Exception {
	    		try {
	    			return new HanaJdbc().querySQLForList(PropertyPlaceholderUtil.getProperty("UNIT.CODE"), null, false);
	    		}finally {
	    			countDownLatch.countDown();
				}
	    	}
		});
	    
	    countDownLatch.await();
	    map.put("COMPANY.CODE", outputDatas.get());
	    map.put("UNIT.CODE", outputDatas2.get());
	    return map;
	}

1.如果outputDatas和outputDatas2的执行时间分别是2秒的/3秒，那么这个方法的最终执行时间是3秒多一点点，

2.如果两者之间没有任何数据上的前置作用，那么可以这样写，性能可以大幅提升。