This chapter focuses on improving the socket server performance using a few useful techniques. Unlike the previous chapter, here we consider multiple clients that will be connected to the server and the communication can be asynchronous. The server does not need to process the request from clients in a blocking manner; this can be done independently of each other. If one client takes more time to receive or process data, the server does not need to wait for that. It can talk to other clients using separate threads or processes.

In this chapter, we will also explore the select module that provides the platform-specific I/O monitoring functions. This module is built on top of the select system call of the underlying operating system's kernel. For Linux, the manual page is located at http://man7.org/linux/man-pages/man2/select.2.html and can be checked to see the...

You have decided to write an asynchronous Python socket server application. The server will not block in processing a client request. So the server needs a mechanism to deal with each client independently.

Python SocketServer class comes with two utility classes: ForkingMixIn and ThreadingMixIn. The ForkingMixIn class will spawn a new process for each client request. This class is discussed in this section. The ThreadingMixIn class will be discussed in the next section. For more information, you can refer to the relevant Python 2 documentation at http://docs.python.org/2/library/socketserver.html and Python 3 documentation at https://docs.python.org/3/library/socketserver.html.

Perhaps you prefer writing a multi-threaded application over a process-based one due to any particular reason, for example, sharing the states of that application across threads, avoiding the complexity of inter-process communication, or something else. In such a situation, if you want to write an asynchronous network server using SocketServer library, you will need ThreadingMixIn.

By making a few minor changes to our previous recipe, you can get a working version of socket server using ThreadingMixIn.

Launching a separate thread or process per client may not be viable in any larger network server application where several hundred or thousand clients are concurrently connected to the server. Due to the limited available memory and host CPU power, we need a better technique to deal with a large number of clients. Fortunately, Python provides the select module to overcome this problem.

We need to write an efficient chat server that can handle several hundred or a large number of client connections. We will use the select() method from the select module that will enable our chat server and client to do any task without blocking a send or receive a call all the time.

Python's select module has a few platform-specific, networking event management functions. On a Linux machine, epoll is available. This will utilize the operating system kernel that will poll network events and let our script know whenever something happens. This sounds more efficient than the previously mentioned select.select approach.

Let's write a simple web server that can return a single line of text to any connected web browser.

The core idea is during the initialization of this web server, we should make a call to select.epoll() and register our server's file descriptor for event notifications. In the web server's executive code, the socket...

Sometimes you need to write a large custom networking application that wants to avoid repeated server initialization code that creates a socket, binds to an address, listens, and handles basic errors. There are numerous Python networking libraries out there to help you remove boiler-plate code. Here, we can examine such a library called Diesel.

Diesel uses a non-blocking technique with co-routines to write networking severs efficiently. As stated on the website, Diesel's core is a tight event loop that uses epoll to deliver nearly flat performance out to 10,000 connections and beyond. Here, we introduce Diesel with a simple echo server. You also...