A Heuristic for Scheduling of Uniform Parallel Processors

Abstract

Parallel processing is one of the arising concepts that are used to process batch of 'n' numbers of tasks through 'm' numbers of parallel processors. These processors are either identical or uniform or unrelated in nature. Scheduling is the discipline of decision making for executing the tasks through parallel processors. One of the important criteria of good scheduling is the schedule with minimum Waiting Time variance (WTV). Therefore minimizing WTV is an objective of designing schedulers. Minimizing the variance of the tasks waiting time in the scheduler's ready queue is a NP-Hard problem. Achieving the Quality of Service (QoS) in the parallel Processors by minimizing the variance of the tasks waiting time in the scheduler's ready queue is a problem of task scheduling. To enhance the performance of parallel processing, it is required to develop a stable and none overlap scheduler where variance of waiting time of individual tasks should be minimized treated as an important parameter. This paper presents a unique task distribution style among uniform parallel processor and then applies the three heuristic based solutions named as VS, BS and RSS are intended for Q_m |prec|WTV minimization problem. The experimental results are compared with the existing approaches and shows that the proposed approach provides the best performance for the compared approaches and problems tested with large task set. The findings of intended algorithm are shown in the form of graph for consonant problems.