A preliminary study of IO model

1. What is? IO

The operating system is responsible for computer resource management and process scheduling . Applications need to go through the operating system , To do some special operations , Such as disk read / write , Memory read / write, etc .

Applications write data to disk , It can only be opened by calling the operating system API To operate .

Applications in user space , There is no real io The process , real io Execute on the operating system , The application initiates once IO The operation consists of two phases ：

• IO call ： Applications make calls to the operating system kernel
• IO perform ： The operating system kernel is complete IO operation

The operating system kernel is complete IO The operation also includes two processes ：

• Data preparation stage ： Kernel wait I/O Equipment preparation number data
• Copy data phase ： Remove data from the kernel buffer , Copy to user process buffer

Actually io Is to transfer the internal data of the process to the external device , Or migrate the data of external devices to the inside of the process . External equipment generally refers to Hard disk ,socket Communication network card . A complete io The process includes these steps

1. The application initiates to the operating system io Call request
2. The operating system prepares data , hold io Data from external devices , Load into kernel buffer
3. The operating system copies data , About the data in the kernel buffer , Copy to user process buffer

3. IO Model

3.1 Blocking io

Know what is io 了 , What is blocking io Well ？

Application process initiation io call , If the kernel data is not ready , The application process has been blocking and waiting , Wait until the kernel data is ready , Copy from kernel to user space , Before returning the success prompt , this io operation , Call it blocking io

shortcoming ： Kernel data is never ready , The user process has been blocked , Waste performance ,

3.2 Non blocking io

Kernel data is not ready , You can first return an error message to the user process , So he doesn't have to wait , Instead, request again by polling , This is non blocking io

Non blocking IO The process is as follows ：

1. Application processes to the operating system kernel , launch recvfrom Reading data .

2. The operating system kernel data is not ready , Return immediately EWOULDBLOCK Error code .

3. Application process polling calls , Continue to launch... To the operating system kernel recvfrom Reading data .

4. The operating system kernel data is ready , Copy from kernel buffer to user space .

5. Completion of invocation , Return to success .

   namely NIO, namely non-blocking io,

   shortcoming ： Relative to blocking io, Although the performance is greatly improved , Frequent polling , Cause frequent system calls , Colleagues consume a lot of cpu resources , You can get zero io Reuse model .

3.3 io Multiplexing

since nio Invalid polling results in cpu resource consumption , When our kernel data is ready , Take the initiative to inform the application process to make system calls again

File descriptor fd File Descriptor, Is a term of computer science , Formally, it is a nonnegative integer , When the program opens an existing file or creates a new file , The kernel returns a file descriptor to the process .

io The core idea of reuse model ： The system provides us with a class of functions （select poll epoll）, You can monitor multiple fd operation . Any one that returns kernel data is ready , The application process is restarted recvfrom system call

3.3.1 Multiplexing select

The application process calls select function , You can monitor multiple fd, Again

3.4 Signal driven io Model

3.5 io Model asynchronous io

3.6 Blocking Non blocking Sync asynchronous io Divide

4 BIO NIO AIO

https://baijiahao.baidu.com/s?id=1718409483059542510&wfr=spider&for=pc