摘要
一个应用程序只有经过分解 ,产生若干任务后才能在元计算系统上调度和运行 .论文提出了基于 DAG的、优化的分代任务调度算法 OGS.该算法为每个 DAG结点指派一个固定优先数 ,根据优先数过滤出互不依赖的任务 ,然后对这些独立任务使用先辈算法进行调度 .OGS算法不仅考虑机器就绪时间 ,还考虑每个任务全部先导的完成时间 .这样可以减少全部参与调度的节点的空闲时间 ,从而优化 Makespan.实验表明 ,OGS算法的性能优于GS和
An application must be decomposed into some tasks (or subtasks) to be executed in metasystems. In this study, an optimal generation scheduling (OGS) algorithm is introduced. First, the algorithm assigns a fixed priority for each task. It, then, filters out all independent tasks according to the same priority. Finally, it maps these independent tasks to all machines available using some heuristics by forerunners for comparison. In OGS Algorithm, expect machine ready time is used, the completion times of all predecessors are also considered as heuristic information. So, idle times in machines can be decreased, and the Makespan then be improved. By test, it is demonstrated that the performance of OGS exceeds GS algorithms.
出处
《计算机学报》
EI
CSCD
北大核心
2002年第6期582-586,共5页
Chinese Journal of Computers
基金
国家"八六三"高技术研究发展计划项目(2 0 0 1AA1110 81
863 -3 0 6-ZT0 1-0 6-2 )资助
西安交通大学行动计划项目资助
