基于Xen的虚拟机动态迁移策略研究

发布时间：2018-03-30 12:41

本文选题：云计算　切入点：虚拟机　出处：《辽宁大学》2013年硕士论文

【摘要】：虚拟化技术采用隔离技术降低了云计算环境下集群的底层异构性，通过虚拟资源设计提高了云计算环境下资源利用率，是云计算的核心技术之一，是云环境下目前的研究热点。虚拟化中的动态迁移技术能够对虚拟机进行无缝迁移，在保证业务运行的同时将虚拟机从源物理主机迁移到目的物理主机，保证迁移的安全性和高效性。目前，云计算环境下的动态迁移性能一般通过SaaS层的具体应用的优化来实现，，这样虽然提高了应用的跨平台适应性，但是却显著降低了动态迁移性能。本文针对Xen平台环境，将实际应用场景和Xen底层的迁移进行结合，提出了基于Xen虚拟机的动态迁移策略。具体工作如下：首先，为了在云计算环境下进行虚拟机迁移，Xen迁移时采用比较传递页位图和跳过页位图的方式来判断内存页是否重传。针对页位图比较带来多次重传增加网络传送开销，提出基于AR模型的内存优化算法，根据记录的内存页所有修改时间间隔来预测内存页的下次修改时间，当下次修改时间大于某个阈值时进行重传。该内存优化算法缩短了虚拟机迁移的时间，减少了虚拟机迁移时的网络开销。其次，在AR模型的基础上，对缩短迁移时间的同时降低业务质量的情况进行分析处理。针对RATE_IS_SAVE模式下迁移时网络带宽最大化分配带来业务预留带宽不足，对原模式进行自适应优化，提出基于指数平滑算法的网络带宽改进策略，依据迭代过程中的脏页产生速率合理预测业务预留带宽，再将剩余带宽分配给迁移，有效缩短了迁移时间，保证了迁移时的业务质量。最后，通过对基于AR模型的内存优化算法、基于指数平滑算法的网络带宽分配策略以及两者结合在多个应用类型下的实验数据收集，分析迁移时间、宕机时间和服务质量等性能指标。结果表明，本文提出的算法在缩短迁移时间的基础上保证了迁移时的业务质量。
[Abstract]:Virtualization technology, which is one of the core technologies of cloud computing, reduces the underlying heterogeneity of cluster in cloud computing environment by isolating technology, and improves the utilization ratio of resources in cloud computing environment through virtual resource design. Dynamic migration technology in virtualization can seamlessly migrate virtual machines and transfer virtual machines from source physical host to destination physical host while ensuring business operation. Ensure the security and efficiency of migration. At present, the dynamic migration performance in cloud computing environment is generally realized through the optimization of specific applications in the SaaS layer, although it improves the cross-platform adaptability of applications. However, dynamic migration performance is significantly reduced. In this paper, a dynamic migration strategy based on Xen virtual machine is proposed by combining the actual application scenario with the Xen underlying migration in the Xen platform environment. The specific work is as follows:. First of all, in order to judge whether the memory page is retransmitted by comparing the passing of page bitmap and skipping the page bitmap, the Xen migration of virtual machine in cloud computing environment brings multiple retransmissions to increase the network transmission overhead. A memory optimization algorithm based on AR model is proposed to predict the next modification time of the memory page according to all the modified time intervals of the recorded memory pages. The memory optimization algorithm shortens the migration time of virtual machine and reduces the network overhead of virtual machine migration. Secondly, on the basis of AR model, this paper analyzes and deals with the situation of shortening migration time and reducing service quality, aiming at the shortage of reserved bandwidth due to the maximum allocation of network bandwidth during migration in RATE_IS_SAVE mode. Based on the adaptive optimization of the original mode, a network bandwidth improvement strategy based on exponential smoothing algorithm is proposed. The reserved bandwidth is reasonably predicted according to the dirty page generation rate in the iterative process, and the remaining bandwidth is allocated to the migration. Effectively shorten the migration time, ensure the quality of the business migration. Finally, through the memory optimization algorithm based on AR model, the network bandwidth allocation strategy based on exponential smoothing algorithm and the experimental data collection under multiple application types, the migration time is analyzed. The results show that the proposed algorithm can guarantee the service quality of migration on the basis of shortening the migration time.
【学位授予单位】：辽宁大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP302

【参考文献】