Completabilefuture of asynchronous tools for concurrent programming

 

Catalog

1： Write in the front

2： The actual combat

1： Test code address ：

2： The test link

3： Comparative data analysis

4： Analysis of other parameters

3： summary

---

1： Write in the front

We are java Whether we will encounter the problem of processing large quantities of data in the development of , If there is such a scene , You check ES Or is it DB, Will return you a large number of data , There may be more than 100000 orders of magnitude , Even millions , Then you take the data and do some processing , Operations may include cpu It's a complex operation , File store , But not limited to the above , What to do with that ？ Also in the use of for Cycle one by one ？ It is estimated that the client will time out , The disgusting thing is that the client has set a timeout , Then pull it again / Many times your ES, Or is it DB, Then at work the next day , The company's network administrator will send you an email , It says “ In the early morning of this morning 3 A little bit later , It lasted about an hour , Something happened on the so and so machine cpu / io It seems a little high , Um.     It is very euphemistic to say .” Then it is time to consider optimization , Whether to use thread pool to solve this problem , Or asynchronous .

2： The actual combat

1： Test code address ：

https://gitee.com/zccpojo/testcomfuture.git

The following two parameters can be adjusted to simulate

2： The test link

The current test parameters are set to ： Number of tasks 2000 individual , The processing time of each task is 10ms

1： Sync ：

Spend total time on 20479ms, And get all the results 2000 strip

2： asynchronous ：

Spend total time on 457ms, And get all the results 2000 strip

3： Comparative data analysis

It's also dealing with 2000 The article number , But we do it asynchronously , I found that the performance improved   20479ms / 457ms  ~ 40 times , What amazing data , Looking at it, I felt a little shivering , However, this is also related to the length of work and the number of tasks , If there are more tasks , The effect will be more amazing .

4： Analysis of other parameters

In the above asynchronous processing mode , We used CompletableFuture, We found that it passed in a thread pool , For information about thread pools, please move to

https://blog.csdn.net/weixin_45191798/article/details/100153524

https://blog.csdn.net/weixin_45191798/article/details/100159321

https://blog.csdn.net/weixin_45191798/article/details/100159830

TIPS： It is best to create a thread pool manually when using a thread pool （ There's a lot of laziness here ）, Be sure to manually close the thread pool after using it , Don't ask why , The question is 《 Alibaba Development Manual 》

3： summary

When doing a lot of data processing , You can consider caching , L2 cache , Even client caching and necessary asynchrony , This article introduces the use of asynchronous + Thread pool to process large quantities of data , Yes, it can be reduced RTT, To improve the throughput of the system . The source code analysis will be updated later , It's not too early , It's gone . Those who are interested can communicate together .