MapReduce计算模式下云资源定量评估方法仿真

摘要

云资源的动态变化和异构性给云资源管理及任务调度带来了巨大的困难,为了准确地掌握云资源动态负载和可用能力信息,需要进行MapReduce计算模式下云资源定量评估.但是采用当前评估方法进行云资源定量评估时,未考虑到应用类型对物理资源需求的特殊性,存在评估误差大的问题.为解决上述问题,提出一种基于马尔可夫的MapReduce计算模式下云资源定量评估方法.上述方法先利用熵增原理筛选出满足MapReduce计算模式下网络用户服务质量和系统最大化的云资源,实现云资源最优调度,保障MapReduce计算模式下网络用户应用需求,并对筛选后的云资源进行动态加权负载评估,对云资源负载过重或是不可用云资源进行迁移,并采用马尔可夫模型对云资源可用能力建模,评估未来阶段云资源的缺失情况对应用服务的影响,依据评估结果对云资源进行调整指导.实验结果表明,所提评估方法有利于实现云资源均衡负载、提高云资源利用率.%This article proposes a quantitative assessment method of cloud resources in computer mode based on Markov's MapReduce.Firstly,this method screened out cloud resources satisfying service quality of network user and maximizing system in MapReduce computer mode by utilizing principle of entropy increase,thus realized optimal scheduling of cloud resources and guaranteed application requirements of network users in MapReduce computing mode,and then carried out load assessment of dynamic weighted for cloud resources screened,transferred overloaded or unavailable cloud resources and modeled available capacity of cloud resources by adopting Markov's model.Finally,this method assessed impact of cloud resources deficiency on application services in the future,according to assessment results,adjusted and guided cloud resources.The simulation results demonstrate that the assessment method in this paper is favorable to realize balanced load of cloud resources and improve utilization ration of cloud resources.