33.pdf

O.E. Oguike et al. evaluate the performance of a shared memory parallel computer system using recursive models. The system is modeled as a single queue parallel server queuing system with a finite single ready queue and C number of processors. The arrival rate of processes is according to Poisson distribution and the service time is according to exponential distribution. The service rate is redefined to represent compute intensive applications where all processors are busy executing a part of a program concurrently. Recursive models are used to evaluate the performance metrics of the system, including the probability density function of the number of processes on the parallel queueing system. The models are simulated using Java programming language and analyzed using statistical regression/trend line analysis. The paper extends previous models by using recursive models to assess the accuracy of the models.