基于请求负载的网格任务模糊控制调度策略

A Dispatch Algorithm for Grid Tasks Based on Request Loads and Fuzzy Control

在面向大规模复杂应用的网格环境中,网格任务的优化调度变得越来越重要。文中首先提出了一种新的网格任务-服务映射算法,即基于任务负载和优先级的网格任务模糊控制调度策略。在考虑到外部网格任务请求负载和优先级的基础上,能够减少众多网格任务的平均完成时间,使得资源利用更加合理,同时还考虑到了网格服务的价格,即在满足任务负载影响和优先级的情况下,选择价格最小的服务。最后,开发了网格任务调度中间件,并在平台上给出了算法与其他几种常见的经典调度算法进行比较的试验结果。实验表明,算法明显优于已有的经典算法,具有较好应用情景。

With the widely use of Grid services, the optimized dispatch of grid tasks becomes more and more important. At first, according to the request load and its priority, a new grid task - service mapping algorithm, also named ＂A Fuzzy Control based Dispatch Algorithm for Grid Tasks＂, is brought out to decrease the average running time of most grid tasks. In this algorithm, the price of grid services is also considered as a fact. And a grid middleware prototype is developed. The algorithm is compared with some other classic algorithms such as random and priority. The experiment shows that it has better performance.