基于云计算环境的虚拟机内存管理研究

发布时间：2018-12-07 15:18

【摘要】：当前随着云计算技术的飞速发展和虚拟化技术的广泛应用，虚拟机内存管理方面的研究越来越受到研究者的重视，相继开发了许多高效的虚拟机内存管理技术。但现有的虚拟机内存管理技术大都集中在对自身内存的调节方面，缺乏了全局状态下对远程虚拟机闲置内存的使用，这就降低了在云计算环境中整个虚拟机集群的内存利用率，也使得虚拟机内存管理技术的应用范围受到限制。本文便是针对当前出现的这些问题展开研究，提出了基于云计算环境的虚拟机内存管理模型DMBMM。 DMBMM主要实现了以下三个技术关键点：虚拟机内存信息管理策略、虚拟机内存平衡策略和基于网络存储的内存切换方法。在虚拟机内存信息管理方面，本文使用了双层内存信息管理结构来管理和统计所有虚拟机的内存利用信息。该方法首先由各个虚拟机集群中的内存监测模块统计自身集群内虚拟机的内存利用信息，然后每个内存监测模块再与内存信息管理服务器交互，完成对所有虚拟机内存信息的统计。在虚拟机内存平衡策略方面，本文根据每个虚拟机集群所处的内存状态是紧缺状态还是充裕状态，分别执行单机平衡策略和多机平衡策略。在基于网络存储的内存切换方法方面，，本文在DMBMM中引入了网络存储技术，利用现有的网络实现了内存的远程使用，解决了现有的虚拟机内存管理技术无法使用远程内存的问题。本文利用MyEclipse开发的两个微基准测试程序获取实验数据。结合本实验软硬件开发、测试的实际环境情况，验证了单机平衡策略和多机平衡策略的有效性。与常用的虚拟机内存管理模型进行对比分析，证明了该模型在有效利用远程内存，提高内存存取速度方面的优势，从实验的角度印证了本论文思路的可行性。
[Abstract]:With the rapid development of cloud computing technology and the wide application of virtualization technology, researchers pay more and more attention to the research of virtual machine memory management, and many efficient virtual machine memory management technologies have been developed one after another. However, most of the existing virtual machine memory management technologies focus on the adjustment of their own memory, and lack of the use of idle memory in the global state, which reduces the memory utilization of the entire virtual machine cluster in the cloud computing environment. It also limits the application of virtual machine memory management technology. In this paper, we present a virtual machine memory management model (DMBMM.) based on cloud computing. DMBMM mainly implements the following three key points: virtual machine memory information management strategy, virtual machine memory balance strategy and memory switching method based on network storage. In the aspect of virtual machine memory information management, this paper uses the double layer memory information management structure to manage and statistics the memory utilization information of all virtual machines. Firstly, the memory monitoring module in each virtual machine cluster counts the memory utilization information of the virtual machine in its own cluster, and then each memory monitoring module interacts with the memory information management server. Complete the statistics of all virtual machine memory information. In terms of virtual machine memory balance strategy, according to whether the memory state of each virtual machine cluster is scarce or abundant, the single machine balance strategy and the multiple machine balance strategy are implemented respectively. In the aspect of the memory switching method based on network storage, this paper introduces the network storage technology into DMBMM, realizes the remote use of memory by using the existing network, and solves the problem that the existing virtual machine memory management technology can not use remote memory. In this paper, two microbenchmark programs developed by MyEclipse are used to obtain experimental data. Combined with the software and hardware development of the experiment, the effectiveness of the single machine balancing strategy and the multi-machine balancing strategy is verified. Compared with the common virtual machine memory management model, it proves the advantage of the model in using remote memory effectively and improving the memory access speed. The feasibility of this paper is verified from the point of view of experiment.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP302;TP333.1

【参考文献】