面向云数据中心的动态高能效虚拟机调度算法研究

发布时间：2018-01-03 23:26

本文关键词：面向云数据中心的动态高能效虚拟机调度算法研究　出处：《西北大学》2017年硕士论文　论文类型：学位论文

【摘要】：云数据中心资源的低效使用已经成为亟待解决的问题,为了解决虚拟机资源管理中高能耗低服务质量问题,本文提出云数据中心虚拟机资源管理策略算法,该策略包含了一系列虚拟机过程管理算法,包括虚拟机最优放置算法、高能效整合过载主机算法、高能效整合轻载主机算法。本文的研究内容,主要包括以下几个方面:(1)为解决虚拟机分配问题,本文严格证明其是NP完全问题,提出虚拟机最优分配算法。该算法结合主机利用率上界阈值及主机性能参数方法对主机的打分结果,确定虚拟机目标放置主机。(2)为解决过载主机检测及整合问题,本文提出高能效整合过载主机算法。此算法利用二次指数平滑法预测虚拟机迁移对预过载主机利用率的影响,进一步确定其上要迁移的虚拟机,消除主机过载风险。(3)为解决轻载主机检测及整合问题,提出高能效整合轻载主机算法。此算法判定选择轻载主机并重分配其上的虚拟机,关闭空闲主机,提高数据中心资源利用效率。(4)为解决虚拟机分配调度框架问题,提出最优虚拟机调度管理算法方案,此算法集合了上述三个组件算法,可高能效地完成虚拟机分配调度全过程。(5)在CloudSim仿真平台中对算法进行实验,结果表明本文提出的云数据中心虚拟机资源管理策略算法与其他解决方案相比,平均减少21.5%的主机能耗,34.2%因迁移导致的虚拟机性能退化情况,70.2%的主机服务水平协议违反时间,68%的能耗服务违反率。
[Abstract]:Cloud data center resources inefficient use has become a problem to be solved, in order to solve the virtual machine resource management in high energy consumption and low quality of service issues, this cloud data center virtual machine resource management strategy, the strategy includes a series of virtual machine process management method, including the virtual machine placement algorithm, high efficiency integrated overload host integration algorithm, energy efficient light load host algorithm. The research content of this paper, mainly including the following aspects: (1) to solve the virtual machine allocation problem, we strictly prove that it is a NP complete problem, a virtual machine optimal allocation algorithm. The algorithm combined with the host using the upper bound of the rate threshold and the performance parameters of the host host the scoring results, determine the target host is placed. The virtual machine (2) to solve the problem of overloading the host detection and integration, this high efficiency integration algorithm. The algorithm of overload and host Prediction of virtual machine migration effect on pre overload host utilization with two exponential smoothing method, further to determine the migration of virtual machine, eliminating the risk of overloading the host. (3) to solve the light load detection and host integration, high efficiency and light load integration algorithm is proposed. This algorithm determines the host selection of light load on the host allocation of virtual machine on the host, idle, improve the utilization efficiency of the data center resources. (4) for solving the scheduling problem of virtual machine allocation framework, and put forward the optimum management of virtual machine scheduling algorithm, this algorithm is a collection of the three component algorithm, can high efficiency to complete the virtual machine scheduling process. (5) experiment of the algorithm in the CloudSim simulation platform, the results show that the proposed cloud data center virtual machine resource management strategy algorithm compared with other solutions, an average reduction of 21.5% in energy consumption due to the migration of host, 34.2% Guide The resulting virtual machine performance degradation, 70.2% host service level protocol violation of time, 68% of the energy consumption service violation rate.

【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP302;TP308

【相似文献】