Page 68 - DCAP403_Operating System
P. 68
Unit 4: Process Management
To implement parallel and concurrent mechanisms you need to use specific primitives of our Notes
operating system. These must have context switching capabilities, which can be implemented in
two ways, using kernel level threads or using user level threads.
Figure 4.8: Diagram of Kernel Thread
Thread Process
User space
Kernel space
Process table
Thread table
If I use kernel level threads, the operating system will have a descriptor for each thread belonging
to a process and it will schedule all the threads. This method is commonly called one to one. Each
user thread corresponds to a kernel thread.
There are two major advantages around this kind of thread. The first one concerns switching
aspects; when a thread finishes its instruction or is blocked, another thread can be executed. The
second one is the ability of the kernel to dispatch threads of one process on several processors.
These characteristics are quite interesting for multi-processor architectures. However, thread
switching is done by the kernel, which decreases performances.
Figure 4.9: Diagram of User Thread
Process
Thread
User space
Kernel space
Run-time Thread table
system Process table
User level threads are implemented inside a specialized library that provides primitives to handle
them. All information about threads is stored and managed inside the process address space.
This is called many to one, because one kernel thread is associated to several user threads. Its has
some advantages: The first is that is independent of the system, thus, it runs faster than context
LOVELY PROFESSIONAL UNIVERSITY 61
