Load balancing real-time periodic task scheduling algorithm for multiprocessor enviornment

Abstract

With phenomenal growth of distributed real time control system in high speed applications, the usage of multiprocessor system has become essential. In order to obtain performance benefits from the resources available in multiprocessor system, real time task have to coordinate and share resources to their maximum use. Load balancing amongst processors of multiprocessing system is one of the most important issues for task scheduling. The processors working on balanced load perform better compared to unbalanced load in terms of response time and resource utilization. Heat dissipation on individual processor is also reduced. In this paper, we propose an algorithm named LBPSA (Load Balanced Partitioning and Scheduling Algorithm) that offers load balancing amongst processors. The algorithm uses an upper threshold, a lower threshold and average utilization of processors for load balancing. Evaluations show that it offers good and almost stable load mean square error under all load conditions. Further the concept is enhanced to generalize it, so that it can be applied on other partition algorithm for incorporating load balancing.