云计算中面向能耗的虚拟机迁移研究

发布时间：2018-12-15 09:28

【摘要】：云计算作为一种新的服务模式理念,在实际生活中,体现为由数百万台的各种物理设备组成的数据中心。云计算利用虚拟化技术,向用户提供虚拟化资源,用户可以随时随地按需请求资源和按需付费。然而,随着云计算数据中心规模的不断扩大,数据中心消耗的能源每年都在成倍地增长,如何通过虚拟机迁移降低整个数据中心能耗这一问题,成为了许多人的研究重点。虚拟机迁移可以分为三个主要过程:主机负载检测,虚拟机选择和虚拟机放置。因此,可以考虑如何针对虚拟机迁移的几个过程进行优化,从而减少数据中心产生的能耗。论文首先针对虚拟机选择,提出了基于性能满足度的虚拟机选择算法,该算法计算出负载主机上每台虚拟机的性能满足程度,从而挑选出性能满足度最低的虚拟机作为待迁移虚拟机;然后针对虚拟机迁移中的虚拟机放置过程,提出了基于最小相关系数的虚拟机放置算法,该算法为待迁移虚拟机挑选出一个迁移后对该主机影响最小的主机。然而,该算法只考虑了CPU利用率。因此,针对最小相关系数的虚拟机放置算法的不足之处,论文提出了改进的虚拟机放置算法:基于预测亲和度的虚拟机放置算法。该算法的特点是不受资源维数的限值,不仅可以考虑CPU利用率,还可以考虑内存,网络等因素。此外,算法引入了一种新的概念:亲和度,它被用来量化虚拟机和虚拟机之间、虚拟机和物理主机之间的关系,从而该算法就可以根据虚拟机和主机之间的亲和度来选择目标主机。最后,基于OpenStack平台实现了虚拟机迁移系统,且在系统中实现了上述的虚拟机选择算法和虚拟机放置算法。实验表明,上述算法在降低能耗方面有一定的成效。
[Abstract]:Cloud computing, as a new service model, is embodied as a data center composed of millions of physical devices in real life. Cloud computing uses virtualization technology to provide users with virtualized resources. Users can request resources and pay on demand anytime, anywhere and on demand. However, with the expansion of cloud computing data center scale, the energy consumed by the data center is increasing exponentially every year. How to reduce the energy consumption of the whole data center through virtual machine migration has become the research focus of many people. Virtual machine migration can be divided into three main processes: host load detection, virtual machine selection and virtual machine placement. Therefore, we can consider how to optimize several processes of virtual machine migration to reduce the energy consumption generated by the data center. Firstly, a performance satisfaction based virtual machine selection algorithm is proposed for virtual machine selection, which calculates the performance satisfaction of each virtual machine on the load host. Thus, the virtual machine with the lowest performance satisfaction is selected as the virtual machine to be migrated. Then a virtual machine placement algorithm based on the minimum correlation coefficient is proposed for the virtual machine migration. The algorithm selects a host with the least impact on the host after migration. However, the algorithm only considers CPU utilization. Therefore, in view of the shortcomings of the minimum correlation coefficient virtual machine placement algorithm, this paper proposes an improved virtual machine placement algorithm: virtual machine placement algorithm based on predictive affinity. The characteristic of this algorithm is that it is not limited by the dimension of resources, so it can not only consider the utilization of CPU, but also consider the factors of memory, network and so on. In addition, the algorithm introduces a new concept: affinity, which is used to quantify the relationship between virtual machines and virtual machines, and between virtual machines and physical hosts. The algorithm can select the target host according to the affinity between the virtual machine and the host. Finally, the virtual machine migration system is implemented based on OpenStack platform, and the above virtual machine selection algorithm and virtual machine placement algorithm are implemented in the system. The experimental results show that the algorithm is effective in reducing energy consumption.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP302

【相似文献】