[swoole series 3.1] have you been asked about processes, threads, and collaborations during the interview?

process 、 Threads 、 coroutines , Have you been asked about the interview ？

Finally started , People are most concerned about , It's also a problem we often encounter in interviews . I believe many students will have experienced this question raised by the interviewer . If your major is computer related , Then the problem is very simple , But if you're not , Then it's really necessary to make up lessons .

Process related issues , In computer major, it is generally explained in the operating system . But I also said in the course related to data structure before , I'm not a computer major , So , This question is really confused for me before . Through these years, I have learned some knowledge related to the operating system , Have some understanding of , And it is really limited to the level of understanding , There is no particularly in-depth study and experiment , So please correct all kinds of mistakes in today's content .

process

process , The simplest understanding is the software we run . It is the execution process of a program . for instance , you are here Windows It's on QQ , Then go to the task manager and you can see a QQ.exe The program of is being executed . This thing is called process .

The process starts and ends , We shut it down QQ , natural QQ The corresponding process is over . For the server operating system , Our service applications are generally not restarted , But it does have the possibility of ending . For example, upgrade the server , Modify the configuration file and so on .

Drive more QQ , Or just like when we played games, we played more games , In fact, it is the application of multiple processes . in addition , Process also has the concept of sub process , It calls the operating system fork() Function created , Generally, such a child process is a copy of the parent process .Swoole The core high concurrency implementation of is actually based on this form of multiprocess , We'll talk about it later .

about CPU Come on , It will use polling or other scheduling algorithms to constantly execute different processes in time slices , So different processes will switch to each other . If you listen to songs , Look at the webpage , It's written in code , It seems to be going on at the same time , Actually CPU At the bottom, these processes are constantly switched and executed , It's just that the speed is too fast for you to feel . The operating system is mainly responsible for the management of these processes 、 Dispatch 、 Switch .

When you use PHP The command line executes a php When you file , In fact, it is also the same as opening QQ Start a PHP process , Or running a PHP Program . At this moment in ps -ef perhaps top In command , You can see the program information you are running . The process will occupy CPU and Memory And so on , According to the operation of the program , It may also produce Networks IO、 Hard disk IO, All of this , Are scheduled through the operating system . The legendary optimization is also aimed at optimizing the utilization of these resources .

Conventional PHP-FPM It is in the form of multi process , It means FastCGI Process Manager .Nginx In fact, it is also multi process , When a program in the same process has time-consuming operations , It's blocking ,Nginx Will wait , therefore , When the request volume is very large ,PHP-FPM Will be very tired . meanwhile ,PHP It's the dynamic language we talked about before , The loading process is complete overall loading , This will also bring loss of performance . Tradition PHP Cannot resist high concurrency , That's why high concurrency performance is poor .

Although there is a problem , But the process is the beginning of a program , If you've done C or Java The beginning of , There must be a main() Method is very important . Actually , stay PHP in , There is also this main() Method , It's just at the bottom C in , We just can't see . No process , The program does not exist , This is the foundation. , It is the root of the following threads and coroutines .

Multi process operation is actually to make full use of multi-core CPU The advantages of , One CPU To correspond to the operation of a process , So as to improve the execution efficiency of the process . Let's say we have Swoole Can be set in worker_num , Its recommended setting value is based on CPU The number of nuclear . similarly ,Nginx The configuration of the worker_processes It is also recommended to configure CPU The number of nuclear .

Threads

Threads , It is the smallest unit that the operating system can schedule operations . It is included in the process , Is the actual operation unit in the process . A thread refers to a single sequential control flow in a process , Multiple threads can be concurrent in one process , Each thread performs different tasks in parallel .

Let me put it this way , The process is a contractor , Thread is its worker . If there is only one contractor , And lay the floor , Brush the wall again , You can only do it step by step . If one of them has a problem , The later work will wait for the completion of the previous work , The operation process in this case is called Serial operation . This is the time , The contractor employed two workers , Let them work at the same time , One floor , One brushes the wall , Is it true that our engineering speed has improved a lot at once , This is called parallel operation .

To sum up , Thread is the younger brother of process , And you can perform some parallel operations synchronously in a process , Some complicated IO operation 、 Businesses that are easily blocked can be put into threads . Think about it , Our program execution efficiency will have a qualitative leap . however , Threads are not as beautiful as they seem , Threads compete for resources , Especially for the same resource . For example, we need to operate a file , Two threads access and modify in parallel , Then it's easy to have problems . This problem is actually very similar to us MySQL Transaction problems in .MySQL How did it work out ？ With a lock , Use transaction isolation level . similarly , In the application , We can also lock . It's like Java Medium synchronized Synchronization lock . After adding the synchronization lock , The thread accessing the lock related content will be blocked , Only when the front one is released can the rear one come in and execute .

Threads are independent of each other , But they can share the resources of the process , There is no way to share resources among multiple processes , Only with the help of external forces or interprocess communication . In fact, this is also very easy to understand , Thread is the younger brother of process , Threads in the same process can get the information of big brother . meanwhile , Threads also consume system resources , Too many threads are the same as too many processes , Give Way CPU 、 Memory And other resources are very hi .

The English name of thread is Thread , stay Java in , You can directly instantiate an object with such a name to perform thread operations . But in PHP in , I guess you haven't seen it . Actually , stay PHP in , There is also a pthreads Expand , There is one in this extension Thread Class can use . however ,pthreads The extension has stopped maintenance . stay PHP7.2 And then there was a parallel Extensions can replace pthreads Expand . Let's talk more about this expansion later , After all, there is no ready-made information , And the documents are all in English .

stay Swoole in , In fact, we can't create threads .Swoole There will be a series of threading capabilities by default , But it is used in the underlying processing network IO And so on . Including the asynchronous tasks we talked about before , In fact, it is just sub process and asynchronous IO Application of process , It's not a thread . stay Swoole in , What we use more is actually what we want to talk about below coroutines .

coroutines

coroutines , Officially speaking , Neither process nor thread , It is more like a function call with no return value . however , If it's more popular , It's not a problem if you think of it as a thread , It's just that this thread is user mode . What does that mean ？ Processes and threads are controlled by the system , Including switching scheduling , And the collaborative process can be operated by itself , No operating system involvement required , The cost of creating destruction is very low . But at the same time , It's not like multithreading can make full use of CPU . stay Swoole in , We need to make use of multi-core with the help of multi-process model in the cooperation process CPU The advantages of .

Come back , A coroutine is a lighter thread , Or let's further say that a coroutine is the younger brother of a thread . The more thread processes open , The more resources are used , The operating system brings more switching consumption . And the coroutine runs on the thread , When a collaboration process is completed , Give up voluntarily , Let another coroutine run on top of the current thread , Reduce thread switching . Empathy , We don't need to open so many threads . Be careful , The synergy here is Concurrent , It's not what the above thread said parallel . Parallelism can really run together at the same time , Concurrency is started together , But still rely on CPU Quickly switch to execute , It's still serial in nature , It just looks like it's being executed at the same time . about parallel and Concurrent Little friends who don't know much can check the relevant information in detail .

For example , So we need to drive 1 A process ,100 Threads , Now use the synergy , We can drive 1 A process ,100 Threads , Then run on each thread 100 Collaborators cheng . Now? , You are equivalent to 10000 A little brother dealing with affairs is working , Have a good time . The main thing is , The machine configuration does not need to be changed , Concurrent synchronization from 100 Turned into 10000 .

Of course , It's not that Xie Cheng is invincible . The coordination process is only waiting IO In order to reuse threads . A thread is waiting for time-consuming IO When operating, it will block , In fact, only threads are known to the operating system at this time , When it's blocked , The coroutine in this thread is also in blocking state . If there is no support for asynchronous calls , In fact, Xiecheng can't play any role . What is asynchronous invocation ？ Namely asynchronous IO ,epoll Have you ever known ？ Or say ,JavaScript The cycle of events in .

Since it's a cycle , In fact, you must also know , Let's summarize . A coroutine is just a thread , Control yourself by yourself , To enable multiple coroutines to be executed in one thread , When the previous process is blocked , Through asynchrony IO Put it there to execute , Then switch to other coroutine operations . At this time , The operating system does not need to switch or innovate new threads , Reduces thread switching time . Be careful , Only threads are truly parallel , Or two unrelated processes , And cooperation , It is not processed in parallel , In the thread , It's also in CPU Switching execution under the time slicing mechanism . A coroutine runtime in a thread , Other processes are suspended . Its switching cost is very low , And it is user controllable . Synergy and process 、 Thread is not a dimensional concept at all , As it says , It's just a function .

At present, collaborative process is a relatively popular development trend , especially Go There are also reasons for the language fire, which naturally supports collaborative processes .Swoole In fact, I have supported the cooperative process from a very early time , And most importantly , We can still use the original traditional way to write our code by using coprocessing , It greatly facilitates our business migration . We will come into contact with these later .

Three way features

Let's summarize the characteristics of these three processes . One by one .

process

• It's the program we execute . There is no shared content between programs , Are their own independent memory space .

• Scheduling switching is completed by the operating system , Users have no perception , There are many switching contents , Spending big , Low efficiency .

• If communication is needed, it is usually through signal transmission , Or external tools .

Threads

• The younger brother under the process , Multiple threads in the same process share memory .

• True parallel execution , You can use CPU The number of nuclear .

• Scheduling switch is smaller than process , But the same is done by the operating system , Moderate overhead and efficiency .

coroutines

• Little brother of thread , But it's more like a function .

• Serial running on threads , Depends on threads , And depends on asynchrony IO , without asynchronous IO Support for , It is no different from threads .

• More lightweight , User mode switching , The cost is low , Very efficient .

summary

Today is all conceptual , There are also many references , Here is not a list , Anyway, it's Baidu , In addition, it also includes some I have learned before Harbin Institute of technology operating system Li Zhi jun Some notes from the professor's class . You can learn about it , There are many good resources in Mu classes in Chinese universities . I have always said before , operating system 、 data structure 、 The Internet 、 Design patterns Four pieces are the four magic weapons of our programmers , Not to mention becoming a great God , But at least understand . otherwise , Your progress will be slow , It's also very limited .

in addition , It's like the beginning , My level is limited , Anything wrong , Please also leave a message to correct , The revised content will also be republished to my blog at any time , Those who need the latest information can go to the blog , There are indeed updates , Wechat will also revise in the message room , So as not to mislead others .