Brief introduction to Linux network I / O model (picture and text)

1. Introduction

The Linux kernel treats all external devices as a file (everything is a file). Reading and writing to a file will call the system command provided by the kernel and return a file descriptor (FD). Reading and writing to a socket will also have a corresponding descriptor, called socket FD (socket file descriptor). The descriptor is a number that points to a structure in the kernel (file path, data area and other attributes).

According to the classification of I / O models by unix network programming, UNIX provides five I / O models.

1.1 blocking I / O model

The most commonly used I / O model. By default, all file operations are blocked.

For example, the socket interface under the I/O model: calling recvfrom in the process space, its system call is returned until the packet arrives and is copied to the buffer of the application process or when the error occurs.

The process is blocked from the beginning of calling recvfrom to its return, so it is called blocking I / O model.

Illustration:

1.2 non blocking I / O model

When recvfrom from the application layer to the kernel, it directly returns an ewouldblock error. Generally, it polls the non blocking I / O model to check this state to see if there is data coming to the kernel.

Illustration:

1.3 I / O reuse model

Linux provides select / poll. The process blocks the select operation by passing one or more FDS to the select or poll system call. In this way, select / poll can help us detect whether multiple FDS are ready.

Select / poll is to sequentially scan whether FD is ready, and the number of FD supported is limited, so its use is restricted.

Linux also provides an epoll system call. Epoll uses event driven mode instead of sequential scanning, so it has higher performance. When FD is ready, the callback function is called back immediately.

Illustration:

1.4 signal driven I / O model

First, turn on the socket interface signal driven I / O function, and execute a signal processing function through the system call sigaction (the system call returns immediately, and the process continues to work without blocking). When the data is ready, a sigio signal is generated for the changed process, notify the application to call recvfrom to read the data through the signal callback, and notify the main loop function to process the data.

Illustration:

1.5 asynchronous I / O

Tell the kernel to start an operation, and let the kernel notify the process after the whole operation is completed (including data replication).

The signal driven I / O model informs when an I / O operation can be started. The asynchronous I / O model has a kernel to inform when the I / O operation has been completed.

Illustration:

2. I / O multiplexing technology

In I / O programming, multithreading or I / O multiplexing technology can be used to process multiple client access requests.

As mentioned earlier, I / O multiplexing technology multiplexes multiple I / O blocks to the same select block, so that the system can process multiple client requests at the same time in the case of single thread.

Compared with the traditional multithreading model, the biggest advantage of I / O multiplexing is that the system overhead is small. The system does not need to create new additional threads or maintain the operation of these threads, which reduces the maintenance workload of the system and saves system resources.

Main application scenarios:

System calls supporting I / O multiplexing mainly include select, pselect, poll and epoll.

Epoll is currently recommended, with the following advantages:

3. Network IO programming in Java

If you are just doing java development, you only need to understand the above content, and you don't have to go deep (just talk about it).

A special article has been published to introduce: summary of Java Network IO programming (bio, NiO and AIO all contain complete example codes)