针对当前网格资源调度系统扩展性差的问题,提出了基于非集中调度模型的以保守Backfilling算法为核心的网格资源调度算法。探讨了网格任务在站点处理器数目异构情况下,网格作业多站点协同调度问题。实验仿真证明,在跨网格站的资源调度中,运用资源预留策略和多站点作业分块可以改善作业的平均响应时间,并起到负载平衡的效果。
In order to improve the scalability of grid scheduling system, this paper presented a scheduling algorithm which was based on decentralized scheduling model. This algorithm, centered on Backfilling strategy, discussed the problem of grid resources scheduling across multi-sites under the condition of heterogeneity of processor amount. Experiment reveals that the scheduling algorithm can improve task turn-around time and load balance effectively.