Processes and threads each have their place in multi-programming, generally to hide latency and to maximize CPU utilization. With the continuing spread of multi-core processors in personal computers, threads are becoming more and more important every day. There are now threads in almost every type of application, including client applications, not just server applications. Soon, there are likely to be multi-core processors even on cellphones, and there will be threaded applications on cellphones.
These multiple cores make it possible to run several different lines of processing at the same time, allowing the computer to runmuch faster than usual; because of this, however, programmersmust now explicitly make use of multi-threaded programming. Unfortunately, parallel programming is very confusing and error prone.
In general, parallel programming can be implemented either by using several concurrent pro- cesses, or by using threads. While processes have each their own address space (separate program counters, heaps, stacks, etc), threads share their address spaces; the programmer can use either of these to obtain concurrency. Besidesmaximizing CPU utilization, the use of parallel programming also helps to hide latency (e.g. waiting for the disk while using the CPU) and to handle multiple asynchronous events.
As a rough analogy, different processes are like different housing subdivisions: quite separate and protected from each other, and with communication possible but relatively dif?cult. In this analogy, different threads within the same process are like roommates:
communication and sharing is easy, but there's much less protection from each other's mistakes.