A common concept in computer science is multitasking. When multitasking, an operating system simply switches between different processes at high speed to give the appearance that these processes are being executed simultaneously, even though it is usually the case that only one process is executing on one single central processing unit (CPU) at any given time. In contrast, multiprocessing is the method of using more than one CPU to execute a task.

While there are a number of different uses of the term multiprocessing, in the context of concurrency and parallelism multiprocessing refers to the execution of multiple concurrent processes in an operating system, in which each process is executed on a separate CPU, as opposed to a single process being executed at any given time. By the nature of processes, an operating system needs to have two or more CPUs in order to be able to implement multiprocessing tasks, as it needs to support many processors at the same time and allocate tasks between them appropriately.

This relationship is shown in the following diagram:

We have seen in Chapter 3, Working with Threads in Python, that multithreading shares a somewhat similar definition to multiprocessing. Multithreading means that only one processor is utilized, and the system switches between tasks within that processor (also known as time slicing), while multiprocessing generally denotes the actual concurrent/parallel execution of multiple processes using multiple processors.

Multiprocessing applications have enjoyed significant popularity in the field of concurrent and parallel programming. Some reasons for this are listed as follows:

• Faster execution time: As we know, when done correctly concurrency always provides additional speedups for your programs, provided that some parts of them can be executed independently.
• Synchronization free: Given the fact that separate processes do not share resources among themselves in a multiprocessing application, developers rarely need to spend their time coordinating the sharing and synchronization of these resources, unlike multithreaded applications, where efforts need to be made to make sure that data is being manipulated correctly.
• Safety from crashes: As processes are independent from each other in terms of both computing procedures and input/output, the failure of one process will not affect the execution of another in a multiprocessing program, if handled correctly. This implies that programmers could afford to spawn a larger number of processes (that their system can still handle) and the chance of crashing the entire application would not increase.

With that being said, there are also noteworthy disadvantages to using multiprocessing that we should consider, as shown in the following list:

• Multiple processors are needed: Again, multiprocessing requires the operating system to have more than one CPU. Even though multiple processors are fairly common for computer systems nowadays, if yours does not have more than one, then the implementation of multiprocessing will not be possible.
• Processing time and space: As mentioned before, there are many complex components involved in implementing a process and its resources. It therefore takes significant computing time and power to spawn and manage processes in comparison to doing the same with threads.