How is process management done in multiprocessing systems?
There are two accost for the process management in multiprocessor systems.
- Master-slave configuration
- Symmetrical multiprocessors
In master-slave configuration one of the processor units is identified as the master processor, which implements the OS routines and performs all control functions in the system. So the master processor decides which process to execute on which processor. All other processors are slaves to the master they basically perform the tasks detailed to them by the master. The disadvantages and advantages of the master-slave configuration are:
- Simplicity
- Poor reliability for the reason that if the master fails the entire system fails.
- Poor scalability that means the load of the master rises with the addition of new slaves.
In symmetrical multiprocessors, a processor is identical in their ability to perform control functions. A processor selection is stimulated whenever an interrupt is raised in the system. The chosen processor handles the interrupt. The remaining of the system can continue its operation. This provides improved reliability than master-slave configuration. A disadvantage of the system is that it doesn't fully exploit the potential of multiple processors in the system.