Heuristic Cycle-Based Scheduling with Backfilling for Large-Scale Distributed Environments
The paper is devoted to comparing the results of an independent job batch scheduling in terms of a virtual organization policy and available resources usage efficiency in large distributed environments like utility Grid. A hybrid approach is proposed on the basis of a cyclic scheduling scheme and backfilling combination. Additionally the paper offers a heuristic shifting procedure which improves jobs execution alternatives selected in the cyclic scheme. The simulation results show that depending on the scheduling efficiency indicator and the level of resource availability each of the approaches is able to provide the best results. Moreover the obtained results are valid under conditions of dynamically varying state of resources and inaccurate user job runtime estimations.