IaaS云中虚拟机的动态迁移技术的研究

发布时间：2018-10-10 16:29

【摘要】：IaaS（Infrastructure as a Service）作为云计算的最底层服务，为架构在其上的各种服务提供支撑功能。这种支撑功能主要是通过对底层物理机进行抽象虚拟化后，借助于虚拟化技术实现的。当今的IaaS随着社会化生产实践，，规模正朝着巨型化的方向发展，如何更好地利用资源、节约能耗已成为当前设计云计算系统的一个关键问题。通过虚拟化技术，可将底层的物理机进行抽象，形成方便管理的集群，然后运用虚拟机动态迁移技术对资源进行高效的监督、管理。在计算和存储模式发生巨大变化的今天，设计实现一个高可用、高可扩展性的云平台，对于满足企业或政府机构对资源频繁变化的需求以及降低构建低碳环保的大型数据中心成本等方面，无疑具有广阔的应用前景。虚拟机的动态迁移是指虚拟机处于对用户提供服务状态，由于某些原因需将虚拟机有源主机迁移到目标主机，但在迁移过程中保持对外服务状态不中断。由于虚拟机在动态迁移过程中要维持服务和迁移两个步骤，所以如何有效地分配带宽一直是该领域的一个难点。本文针对云计算服务IaaS展开研究，基于开源云平台OpenStack实现相应的迁移环境，进而在虚拟机的迁移带宽分配方面做进一步的分析、研究，主要完成以下几个方面的工作：深入分析研究Clark等人提出的虚拟机动态迁移算法，并基于开源云平台OpenStack实现虚拟机的动态迁移环境；基于排队论的M/M/1模型，对迁移代价和迁移带宽进行建模，设计了一种新型的用于衡量虚拟机动态迁移性能的评价标准。通过相应实验对该模型与真实环境中的情形进行分析与评价；在对Clark等人设计的变带宽以及常带宽分配算法分析、研究的基础上，设计一种新的变带宽分配算法，并对其在总迁移时间和停机时间两个因素上与Clark等人的带宽分配算法进行比较分析，得出满意的结论。
[Abstract]:As the lowest service of cloud computing, IaaS (Infrastructure as a Service) provides support for various services. This support function is realized by means of virtualization technology after abstract virtualization of the underlying physical machine. With the practice of socialized production, the scale of IaaS is developing towards mega scale. How to make better use of resources and save energy has become a key problem in the design of cloud computing system. Through the virtualization technology, the underlying physical computers can be abstracted to form a convenient management cluster, and then the virtual machine dynamic migration technology is used to efficiently supervise and manage the resources. At a time when computing and storage patterns have changed dramatically, a cloud platform with high availability and scalability is designed and implemented. It will undoubtedly have a broad application prospect for meeting the needs of enterprises or government agencies for frequent changes in resources and reducing the cost of building large data centers with low carbon and environmental protection. The dynamic migration of virtual machine means that the virtual machine is in the state of providing service to the user. For some reasons, the active host of virtual machine needs to be migrated to the target host, but the external service state is not interrupted during the migration process. Since virtual machines need to maintain service and migrate two steps in the process of dynamic migration, how to allocate bandwidth effectively has always been a difficulty in this field. This paper focuses on the research of cloud computing service IaaS, implements the corresponding migration environment based on open source cloud platform OpenStack, and makes further analysis and research on the allocation of migration bandwidth of virtual machine. The main works are as follows: the dynamic migration algorithm of virtual machine proposed by Clark et al., and the dynamic migration environment of virtual machine based on open source cloud platform OpenStack, the M/1 model based on queuing theory, the dynamic migration algorithm of virtual machine proposed by Clark et al. The migration cost and bandwidth are modeled and a new evaluation standard is designed to measure the dynamic migration performance of virtual machines. Based on the analysis of the variable bandwidth and constant bandwidth allocation algorithms designed by Clark et al, a new variable bandwidth allocation algorithm is designed. Compared with the bandwidth allocation algorithm of Clark et al, the two factors of total migration time and downtime are compared, and a satisfactory conclusion is obtained.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP393.09;TP302

【相似文献】