QoS指导下的网格资源选择方法

在高性能并行分布式网格计算环境下,计算节点具有动态、异构、分布和多管理域特性,这使得网格资源的发现及管理变得比较困难。在对网格调度模型分析的基础上,使用模糊聚类的相关理论提出在任务调度之前对资源进行预处理,依据用户QoS需求权重对机器性能进行聚类,得到和用户QoS需求最匹配的较优资源子集。这样既合理地减少了映射的目标资源,又实现了目标资源子集与用户QoS需求的最优匹配。在一定程度上提高了整个系统的性能,满足了不同用户的特殊需求。

QoS Guided Selection Approach for Grid Resource

Under the high-performance parallel distributed computing of grid environment,the discovery and management of the grid recourse become more difficult because of the compute nodes with dynamic,heterogeneous,distributed and multi-domain management features.This paper based on the analysis of grid scheduling model,using the related theory of fuzzy clustering,presents the resources pretreatment before tasks scheduling.The Algorithm of resources pretreatment is to cluster the machine performance according to the QoS requirement to get resource subsets.It is not only reducing the mapping of target resources but also achieving the optimal match of target resources and user’s QoS requirements.Thus,it meets the needs of different users and increases the performance of overall system to some extent.