THE IMPACT OF MULTIPROCESSOR DISTRIBUTED MEMORY SYSTEMS ON PARALLEL COMPUTING APPLICATIONS

Abstract

Compute-intense calculations and parallel computing concepts on Central Processing Units (CPU) require long execution times. Hence, to overcome this challenge which is as a result of limited memory for processes of high matrix orders, developers nowadays use parallel computing approach to develop applications. The aim of the study is to show the impact of multiprocessor distributed memory systems in executing these applications built using parallel computing approach with shorter execution times. The parallel computing approach allows for the breaking of high matrix order processes into smaller independent parts that will run concurrently or in parallel. In this study the client-server model was used. The algorithm used was able to compute the client and server CPU time, execution time, server CPU’s percentage usage on high matrix orders. It was seen that the average CPU usage for the 4096 matrix order was almost 50% less the time that it usually took a uniprocessor to execute. While for a large matrix order of 5032 it took a little over the same time by the multiprocessors to execute. The study revealed that parallel computing applications execution time can be effectively reduced by the use of multiprocessor distributed memory systems.