Introduction and response to high concurrency

One 、 Introduction of high concurrency

What we call high and pointed out two aspects , The first aspect refers to that there are a large number of requests to access the system at the same time The second aspect refers to that there are a large number of requests to access the system in a period of time , There is an indicator in the general time period , Is the number of requests that can be processed in one second .

Two 、 High concurrency processing

2.1、 Improve the performance of a single server

For example, add memory ,CPU Auditing and so on .

2.2、 Deploy multiple machines to share the pressure

Use the help of load balancing , At the same time, the program needs to support the deployment of load balancing , For example, introduce Redis To unify storage Session;

2.3、 If the capacity of a single machine is limited , Far below the ceiling , You need to find performance bottlenecks ：

2.3.1 database

Use read-write separation 、 Sub database and sub table 、 Multimaster 、 Improve the database processing capacity from the architecture ;

2.3.2 Introduce caching mechanism

Store data in memory , That is, the cache mechanism is introduced , Check the cache first , Query the database if it does not exist in the cache , Then store it in the cache .【 Cache consistency issues ： You can subscribe binlog journal , perhaps MQ Methods to ensure the final consistency of data 】

2.3.3 introduce Redis colony

Redis Although the single machine performance is very high , But in order to support high concurrency , Generally, it needs to be deployed in a cluster .

3、 ... and 、 The solution to the problem of data processing ability

The maximum processing capacity of a business is tenthousand a second , But what should I do if there are millions of requests in a second ？

3.1 Use MQ

First put the request in MQ, Line up at the exit to consume according to the processing capacity , This method may require users to wait for a while to process the results , such as ： Payment scenario 、 Buying scene 、 Seckill scene, etc .

3.2 Current limiting

If the processing capacity of the system is limited , It can only carry 10000 traffic per second , But what about millions of requests , At this time, the current can also be limited , Let the flow enter the system for processing in batches .

3.3 DNS Load balancing

If the request inlet flow is too large , You can use DNS Load balancing , Distribute the same domain name access to different nodes for processing .

3.4 CDN

Put static resources in CDN To speed up access ;

3.5 Search using ES;

3.6 Split into micro Services

Split the system , Change to micro service , Deploy each module as a cluster independently , That is, each module is a cluster .

3.7 other

In terms of code , Use multithreading to improve processing power , Use connection pooling to reduce database connections 、 The overhead of remote calls .