The multiprocess vs the multithreading model

Here are some pros of the MT model over the single-threaded process:

Context	Multiprocess (single-threaded) model	Multithreaded (MT) model
Design for parallelized workloads	Cumbersome Non-intuitive Using the fork/wait semantics repeatedly (creating a large number of processes) isn't simple or intuitive either	Lends itself to building parallelized software; calling the `pthread_create(3)` in a loop is easy and intuitive as well Achieving a logical separation of tasks becomes easy The OS will have threads take advantage of multicore systems implicitly; for the Linux OS, the granularity of scheduling is a thread, not a process (more on this in the next chapter) Overlapping CPU with IO becomes easy
Creation/ destruction performance	Much slower	Much faster than processes; resource-sharing guarantees this
Context switching	Slow	Much faster between the threads of a process
Data sharing	Done via IPC (Inter-Process Communication) mechanisms; involves a learning curve, can be fairly complex; synchronization (via the semaphore) required	Inherent; all global and static data items are implicitly shared between threads of a given process; synchronization (via the mutex) is required

Table 4: Multiprocess versus multithreading model – pros of the MT model

Table of Contents for
The multiprocess vs the multithreading model – pros of the MT model