An enhanced and effective preemption based scheduling for grid computing enabling backfilling technique
Anurag Mishra · 2015
Recent improvements in designing the scheduleing algorithms for cluster mainly focuses on the advance reservations and Backfilling algorithms. Backfilling has been proposed and launched in various flavors to have an idealistic utilization of the resources and computing capacity of the cluster. However backfilling demands the runtime prediction of the job. Preemption of process in the scheduling queue is done to reduce the starvation and to fulfill the requirements of the proposed algorithm. We have applied the proposed algorithm in the cluster designed through TORQUE and preemption support through BLCR. Another implementation is done for MPI jobs in LAM environment with BLCR as a job checkpoint and restart facility. Proposed algorithm is compared with the Backfilling techniques with preemption support. Algorithm showed around 3%-5% reduction in average turnaround time and 6%-8% increment of throughput rate.