Multilevel Queue-Based Scheduling for Heterogeneous Grid Environment

Grid computing is the federation of pooling resources so as to solve large-scale problems. Scheduling is the main issue in grid computing and is the process of making scheduling decisions over multiple grid resources. In this paper, we propose a scheduling technique which classifies the subtasks based on the priority assigned by the user. This is mainly applicable in places where the high priority critical subtasks may need to be scheduled prior to other low priority subtasks. We thus segregate the subtasks into three queues based on their priority. Subtasks within each queue are reordered based on two new parameters, viz. complexity factor and realization factor, with computational complexity defined as the time of computation of a process. We evaluate the realization factor as the product of number of operations per cycle per processor and the speed of the processor. The subtasks are assigned high priority when both complexity factor and realization factor are high. Once the processes are classified into three queues we make use of a technique similar to round robin that reduces starvation of low and medium priority subtasks. The effectiveness of Starvation free (SF) Scheduling algorithm is evaluated through simulation results.