In: Computer Science
what is and characteristics of multiprocessor computer systems
what is and characteristics of multiprocessor computer systems
Multiprocessor system:
Multiprocessor system can be defined as a system architecture, in which two or more processing units, linked together to generate parallel processing and accelerate the execution speed along with other significant objectives like fault tolerance and application matching.
Depending on the functionality multiprocessor system can be classified as nine types.
I. Loosely coupled multiprocessor system
II. Tightly coupled multiprocessor system
III. Homogeneous multiprocessor system
IV. Heterogeneous multiprocessor system
V. Shared memory multiprocessor system
VI. Distributed memory multiprocessor system
VII. Uniform memory access (UMA) system
VIII. cc–NUMA system
IX. Hybrid system
Depending on sharing data within processing units and memory, multiprocessing system cab be broadly classified as two types.
In shared memory multiprocessors, all the processing units share the common memory.
According to the figure, all processing units (PU) share same memory using a common interconnection network.
In distributed memory multiprocessor every processing unit has its own memory and processing unit share data with its own memory unit.
According to the figure, it is found that, every processing unit (PU) has its own memory unit.
Characteristics of Multiprocessor system:
I. Multiprocessor system is basically the interconnection of two or more processing units with memory and i/o devices.
II. This system works on the principal of MIMD architecture, where multiple instructions from different processing units work on multiple data.
III. Fault tolerance is one of the significant characteristics of multiprocessor system, where the failure of any processor can be compensated by any other workable processor.
IV. Multiprocessing can decompose the program into parallel executable tasks and enhances the execution speed of processing.
V. In shared memory approach every processor acts as tightly coupled, where as in distributed approach as every processor has its own private memory, so they show loosely coupled approach.