Swoole collaboration

Introduce

from 4.0 Version start  Swoole  Provided complete coroutines （Coroutine）+  passageway （Channel） characteristic , Bring new  CSP  Programming model .

1. Developers can write code in a synchronous way without perception asynchronous IO  The effect and performance of , It avoids the discrete code logic brought by the traditional asynchronous callback and the code cannot be maintained due to multi-layer callback
2. At the same time, because the bottom layer encapsulates the collaborative process , So compared with the traditional  PHP  Layer collaboration framework , Developers don't need to use  yield  Keywords to identify a collaborative process  IO  operation , So there's no need to talk about  yield  In-depth understanding of the semantics of and the calls to each level are modified to  yield, This greatly improves development efficiency
3. Various types of perfect Process client , It can meet the needs of most developers .

A coroutine can be simply understood as a thread , It's just that this thread is user mode , No operating system involvement required , Very low cost to create destroy and switch , Unlike threads, coroutines cannot take advantage of multiple cores cpu Of , Want to use multi-core cpu Need to rely on  Swoole  The multi process model of .

What is? channel

channel  It can be understood as message queue , It's just a message queue between processes , Multiple processes pass  push  and  pop  Operate production messages and consumption messages , Used for communication between processes . It should be noted that  channel  There is no way to cross process , Only one  Swoole  Communication between processes , The most typical application is Connection pool and Concurrent invocations .

What is a process container

Use  Coroutine::create  or  go  Method to create a coroutine ( Reference resources Alias section ), Only in the created collaboration can we use the collaboration  API, And the process must be created in the process container , Reference resources The coroutine container .

Scheduling

Here we will try to talk about what is collaborative scheduling , First of all, each coroutine can be simply understood as a thread , As we all know, multithreading is to improve the concurrency of programs , The same multi concurrency is also to improve concurrency .

Each request of the user will create a collaboration , After the request, the collaboration ends , If there are thousands of concurrent requests at the same time , There will be thousands of processes in a process at a certain time , that CPU Resources are limited , Which collaboration code is executed ？

Decide whether to let CPU The code decision process of executing which collaboration is Scheduling ,Swoole  What is the scheduling strategy of ？

• First , I found that this line of code encountered  Co::sleep()  Or network  IO, for example  MySQL->query(), This must be a time-consuming process ,Swoole  It will Mysql Connected Fd Put it in  EventLoop  in .

  • And then give up this process CPU For other processes ： namely  yield( Hang up )
  • wait for MySQL After the data is returned, continue to execute this process ： namely  resume( recovery )

• secondly , If the code of the collaboration has CPU Intensive code , Can be opened  enable_preemptive_scheduler,Swoole Will forcibly let this process yield CPU.