Heuristic strategy for multiphase applications scheduling onto identical parallel multiprocessor

Runtime characteristics of most of applications exhibit time varying phase behavior. Different phases of multi-phase applications may require different number of processors in order to execute these phases. These types of problems are referred as multiprocessor scheduling problems which are inherently hard to schedule. The Scheduling strategies which consider applications phase behavior play an important role in achieving high performance and throughput in multiprocessor systems. In this paper, we have proposed two heuristics longest chain maximum processor criticality ratio and maximum criticality first to schedule multiprocessor tasks requiring arbitrary number of processors and having unit processing time with chain precedence constraint. In the proposed heuristics, tasks are scheduled on the basis of maximum criticality and longest chain first with maximum processor criticality ratio. We also compare the performance of both heuristics in terms of make-span time with 0/1 knapsack results. The experimental results have shown for both proposed heuristics. The proposed heuristic reduces the make-span time of the schedule which suggests an improvement in performance of the system.