Server concurrency - note 1

One 、 Improve server concurrency

1. queue + Connection pool

2. The main business logic is processed on the application server , The server only does auxiliary business processing

3. cache , Reduce database access .

3.1 Cache synchronization , When the cache fails, query the relevant data in the database , Poor real-time performance ;

3.2 Once the data in the database is updated , Notify the front end of the cache update , In this way, the real-time performance is better .

3.3 Cache page feed , Out of memory , Swap inactive data out of memory , Common algorithms FIFO,LRU( It is seldom used recently ),LFU( Least frequently used ).

4. Database read and write operations are separated , There are usually more query operations than update requests , Separate read and write operations to achieve load balancing . Databases usually have replication Mechanism ,Master The database is used to update data , Multiple Slave The database is used for query operations , When there is data update, it should be synchronized to multiple Slave database .

5. The application server can also realize load balancing .

5.1 Add a task server to realize , The task server monitors the load of the application server 、CPU Etc , Select the server with the lowest load to distribute tasks .

5.2 The application server takes the initiative to fetch tasks from the task server for processing , This load balancing will be fairer . The constraint is that all application servers can handle the same task .

6. The task server should have 2 One or more ,2 Heartbeat detection between stations , When one of them fails , It should be able to transfer to another one immediately ,FailOver Mechanism .

7. Four killers of server performance ：

7.1 Data copy ---- cache

7.2 Environment switching ---- How to choose between single thread and multi thread , If the server is single core , Using multithreading will increase the switching overhead , The state machine programming mechanism is the most efficient ; If the server is multi-core , Multithreading can give full play to multi-core performance , However, environment switching should be avoided as far as possible , Not more is better .

7.3 Memory allocation — Reduce the amount of memory allocated to the operating system .

7.4 Lock competition — Use logic to avoid lock use .