摘要
针对网格依赖任务重调度所面临的效率低与触发频繁的问题,提出资源动态组织支持的网格依赖任务调度机制.该机制以资源的动态组织为核心,基于资源动态性度量结果对资源进行性能聚类分析,并过滤性能相似资源中的强动态性资源,以在减少资源数量的同时提高重调度备选资源的稳定性.实验表明基于该机制的重调度算法保持了静态调度策略在动态网格环境下相对于动态调度策略的性能优势,从而验证了该机制解决网格依赖任务重调度所面临问题的有效性.
To solve the low efficiency and triggered frequently problem of grid dependent tasks rescheduling, this paper proposes an resources' dynamic organization supported grid dependent tasks rescheduling mechanism, which takes the resources' dynamic organization as the core, and classifies the resources by performance with clustering analysis method based on the resources' dynamic evalua- tion results, and filters the strong dynamic resources from the performance similar resource group, in order to reduce resources while increasing the filtered resources' stability. Experiments show that the rescheduling algorithm based on the proposed mechanism could maintain the performance advantage for the static scheduling strategy compared to the dynamic, thus show that the proposed mecha- nism is effective to resolve the low efficiency and triggered frequently problem faced by grid dependent task rescheduling.
出处
《小型微型计算机系统》
CSCD
北大核心
2009年第12期2389-2397,共9页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60773218)资助
辽宁省博士启动基金(200905212)资助
关键词
网格
依赖任务
重调度
资源动态组织
资源动态性度量
grid
dependent tasks
rescheduling
resources' dynamic organization
resource dynamic evaluation