基于分类挖掘的网格资源分配研究

摘要

Making use of classification data mining algorithm to analyze user' s historical access information, this paper got user' s classification access rules and patterns in cluster renvironment. Firstly, it constructed a classification mining-based resource scheduling model. Secondly, it designed a UA algorithm to allocate all of the users' tasks in each cluster. Finally, it applied a new resource algorithm CDMRA to assign users' tasks to idle CPU resources in every node. Experiments show that CD-MAR algorithm can reduce the resource reallocation times and improve the efficiency and accuracy of resource allocation compared to other algorithms. It can increase the utilization of grid resources.%根据用户访问网格资源的历史信息,采用分类算法对此信息进行挖掘,得出用户使用集群资源的访问规则和模式,在此基础上构造一种基于分类挖掘的资源调度模型、用户调度UA算法以及资源调度CDMRA算法,分别将用户请求调度到各个集群中闲置的CPU资源.实验证明,采用基于分类挖掘的资源分配策略相比其他算法可以减少资源分配过程中对资源的重新分配次数,可以提高网格资源的利用率.