网格工作流动态调度研究

Study of Dynamic Scheduling for Grid Workflow

现有的网格工作流调度算法大都利用遗传算法所具有的并行性和全局解空间搜索的特点来解决工作流调度问题。但是,现有的调度算法没有对动态调度问题进行处理。文中针对网格服务的动态性,提出了服务资源信息中心的概念并给出了网格工作流管理系统的体系结构。在现有的基于遗传算法的网格工作流调度算法的基础上提出了网格服务工作流动态调度算法,补充了不同工作流过程模型的适应度函数的计算。

As an NP problem, grid service workflow（GSF） scheduling is difficult to be solved by means of classic algorithms. Featured in searching concurrently and globally, genetic algorithm can be a better option for solving GSF scheduling problem, so it is widely adopted by present papers. But most of them can not deal with the dynamic scheduling of grid service. Pointing to the problem, this paper proposes the concept of Center of service information and the architecture of service-oriented grid workflow. This paper also supplements the fitness calculation for different workflow model. In the end, the algorithm is proved to have better convergence by the analysis of instances.