A New Heuristic Approach for Scheduling Independent Tasks on Heterogeneous Computing Systems
Marjan Kuchaki Rafsanjani, Amid Khatibi Bardsiri · International Journal of Machine Learning and Computing · 2012
Scheduling is one of the core steps to efficiently exploit the capabilities of heterogeneous computing systems.The problem of mapping meta-tasks to a machine is shown to be NP-complete.The NP-complete problem can be solved only using heuristic approach.There are a number of heuristic algorithms that were tailored to deal with scheduling of independent tasks.Different criteria can be used for evaluating the efficiency of scheduling algorithms.The most important of them are makespan, flowtime and resource utilization.In this paper, a new heuristic algorithm for scheduling meta-tasks in heterogeneous computing system is presented.The proposed algorithm improves the performance in both makespan and effective utilization of resources by reducing the idle time of the machine.The performance analyses show that the proposed algorithm has a better resource utilization rate and reduced makespan than the other known algorithms.