A general model for performance investigations of priority based multiprocessor system

Abstract

A general discrete time semi-Markov model is developed to investigate the effects of task priorities on the system performance of a multiprocessor system with crossbar interconnection network. The number of priority levels associated with the tasks in the system, connection times of different priority level requests, interrequest time, number of processing elements, and the number of shared resources are the parameters involved in estimation of the performance of the system. The bandwidth, queue length at a memory, waiting time for requests at different priority levels, and processor utilization are the performance measures quantified from the analysis. The results reveal the advantage received by the tasks at higher priority levels and the starvation experienced by the lower priority tasks. This information should be useful in the real-time task scheduling, load balancing, and performance optimization. The results obtained are validated with simulation. >