摘要
在尊重网格资源本地调度策略前提下,提出一种基于云模型的动态服务能力评估方法;根据动态性能评估尺度对服务进行聚类,提出了一种基于PSO的自适应的服务动态聚类方法,将提供相同或相似QoS的服务划分到同一个服务簇中,从而缩小了任务调度的问题规模;基于服务动态聚类提出了一种元任务调度算法,理论分析该算法降低了不聚类调度算法的复杂度.实验结果表明本文提出的调度算法在时间复杂度与用户QoS保障方面优于以前提出的调度算法.
Not constrained by local scheduling policy of Grid site, a dynamic service evaluation method based on eloud model is presented. According to dynamic service evaluation, an adaptive and dynamic service clustering method is derived from PSO （Particle Swarm Optimization ）-based clustering algorithm. It gathers the services with similar or same QoS （Quality of Service） into one cluster. A dynamic recta-task scheduling algorithm, called SCTDSA （Service-Clustering-based meta-Task Dynamic Scheduling Algorithm）, is proposed in light of service clustering; The time complexity of the algorithm is lower than that without service-clustering through theoretic analysis. The experiment results show that our algorithm outperforms prior well-established algorithms in terms of time complexity and user QoS guarantee.
出处
《小型微型计算机系统》
CSCD
北大核心
2009年第2期248-254,共7页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60673165
60433020)资助
关键词
服务网格
性能评估
动态聚类
调度算法
service grid
performance evaluation
dynamic clustering
scheduling algorithm
