摘要
针对云计算环境下大量用户请求同时到来造成的短暂峰值,使得用户的QOS服务需求不能得到有效保证,设计了一种基于MPI并行计算模型和SOM的异构资源聚类方法;首先,设计了一种改进的MPI的树状层次结构模型,然后,定义了基于SOM自组织映射的资源聚类算法,为了提高资源聚类精度,将PSO算法用于SOM的参数优化中,使得SOM在初始时刻就具有一个较好的连接结构;最后,为了充分满足用户请求的QOS需求,将MPI树状层次结构与基于PSO的SOM资源聚类算法相结合,并提出了具体的基于MPI的SOM资源聚类算法;为了验证文中方法,在Matlab仿真环境中进行测试,实验结果表明文中方法聚类精度为100%,且与其它方法比较,具有较高的聚类精度和较少的执行时间,是一种云计算环境下的可行资源聚类方法。
Aiming at the transient peak when the amounts of users appear at the same time in Cloud environment, leading to the QOS of user can not be guaranteed, a heterogeneous resource clustering method based on MPI parallel computing model and SOM is proposed. First- ly, the improved MPI tree hierarchical model is designed, then the resource clustering method based on SOM is defined, in order to improve the accuracy of resource clustering, the PSO algorithm is used to the optimization of SOM parameters and the SOM has a better structure at the initial time; finally, in order tO satisfy the QOS need of users, the MPI tree hierarchical structure is applied in SOM clustering algorithm based on PSO, and the specific algorism based on MPI and SOM is proposed. For verifying the method in this paper, the experiment is simu- lated in the Matlab environment, the experiment shows the method in this paper has the clustering precision 100%, and compared with the other methods, it has the high cluster accuracy and less executing time, therefore, it is a feasible resource clustering method in cloud environment.
出处
《计算机测量与控制》
北大核心
2014年第8期2523-2525,2549,共4页
Computer Measurement &Control
关键词
云计算
并行算法
自组织映射
粒子群
cloud computing
parallel algorism
self organizing mapping
particle swarm