A Reliable Distributed Grid Scheduler for Mixed Tasks

Abstract

AbstractScheduling of jobs is one of the crucial tasks in grid environment. We consider non-preemptive scheduling of mixed tasks in a computational grid. Recently, a general distributed scalable grid scheduler (GDS) was proposed, which prioritizes mission-critical tasks while maximizing the number of tasks meeting deadlines. However, the GDS scheduler did not consider the reliability factor, which may result in low successful schedule rates. In this paper, we propose a novel distributed grid scheduler which considers various parameters - Priority, Deadline, CCR and reliability of grid nodes. The proposed scheduler maintains the tasks allocated to deficient grid nodes in a queue. Further the queued tasks are rescheduled to the other nodes of the grid. It is observed that RDGS-MT scheduler shows a significant improvement in terms of successfully scheduled tasks (hard, firm, soft) as compared to a GDS-S (GDS without shuffle phase). The results of our exhaustive simulation experiments demonstrate the superiority of the proposed scheduler over the GDS-S scheduler.KeywordsGrid ComputingSchedulingRe-SchedulingDistributed SchedulerReliabilityPriorityDeadline