SGER: Task scheduling in computational grids

该SGER提案旨在开发和模拟启发式调度算法，以有效地将任务映射到网格环境中的异构集群系统。本文将采用两种方法进行算法的开发和研究：列表调度和聚类。列表调度方法包括任务与集群组件的智能匹配，列表中关键任务的优先级，以及在适当的地方复制任务副本。在分析重复成本与总体性能收益之间的权衡时，任务重复将是具体研究的主题。将计算已开发计划的性能界限，并使用随机有向无环图（dag）评估计划。

This SGER proposal seeks to develop and simulate heuristic scheduling algorithms to efficiently map tasks to heterogeneous cluster systems in grid environments. Two approaches will be taken to the algorithm development and investigation: list scheduling and clustering. The list scheduling approach involves intelligent matching of tasks to cluster components, prioritizing critical tasks in the list, and duplicating task duplication where appropriate. Task duplication will be the subject of specific study, in analyzing tradeoffs in the cost of duplication against overall performance gains. Performance bounds will be calculated for developed schedules, and the schedules will be evaluated using random Directed Acyclic Graphs (DAGs).