基于约简任务资源分配图的网格依赖任务静态调度

网格资源环境的动态性和异构性对依赖任务的静态调度策略产生了极大的挑战.提出了一个能够更好地适应资源异构性,并完整支持调度、重调度的统一静态调度模型,即基于约简任务资源分配图(RT-RAG)的调度模型和调度算法.基于该模型的调度问题归结为RT-RAG的优化选取问题.实验分析表明,基于该模型的调度方法更好地适应了网格环境的动态性和异构性,使得静态调度策略与动态调度策略相比依然保持着较大的性能优势,并且也优于基于DAG任务图调度模型的AHEFT静态调度算法,特别是在资源差异较大的情况下.

Static Grid Dependent Tasks Scheduling Based on Reduced Task-Resource Assignment Graph

The dynamic and heterogeneous nature of grid resources is a great challenge to static scheduling strategies of dependent tasks.A new static scheduling model based on RT-RAG(reduced task-resource assignment graph) and algorithm are proposed which are more suitable for the grid heterogeneity and support unified scheduling/rescheduling process.The(re)scheduling problem based on the model is thus regarded as an optimization selection problem of RT-RAGs.The experimentation shows that the static scheduling algorithm has better adaptability to grid dynamic and heterogeneity and keeps greatly its superiority up in comparison to the dynamic one and it is also better than the AHEFT algorithm based on the DAG task graph model,especially in case of great resources heterogeneity.