虚拟机内存管理与性能对比测试的研究

发布时间：2018-10-05 09:09

【摘要】：随着计算机行业的蓬勃发展，硬件系统的处理能力愈发强大，而软件系统则稍有滞后，带来的问题就是企业无法充分利用现有的硬件资源，造成计算能力的浪费。虚拟化技术的出现，提供了解决这一问题的一种途径。论文从虚拟化技术应用场景和定义开始，讨论了目前应用最广泛的三种VMM模型及其对应的虚拟化方案，总结了x86架构上的这三种不同方案的优缺点。随后介绍了主流半虚拟化虚拟机Xen的架构和相关技术。以Xen的内存管理为重点，，结合源代码，研究了用于物理内存管理的气球驱动算法，并对其在多虚拟机运行下存在的页面抖动问题进行了分析,并针对该问题，提出了“集中式判断气球驱动算法”，给出了算法的流程图和伪代码，且对使用修改后算法的虚拟机的进行了性能测试分析。进而为了能够合理、有效、客观地评估一个经过改进后的虚拟化计系统，作者结合层次分析法和离差最大化法，提出了主客观结合的“三层系统性能评价模型”，并利用了优化过内存管理算法的最新版Xen Client2.1和VMware Workstation作为待评估对象，对这两款桌面虚拟化系统进行了综合的性能评价。最后，论文介绍了“集中式判断气球驱动算法”和“三层系统性能评价模型”的进一步改进思路和发展前景。论文工作可对Xen虚拟机系统的内存管理研究和计算机系统性能评价的研究提供相关参考，同时对其他系统的分析设计也具有一定参考价值。
[Abstract]:With the rapid development of computer industry, the processing power of hardware system becomes more and more powerful, while the software system lags behind. The problem is that the enterprise can not make full use of the existing hardware resources, resulting in a waste of computing power. The emergence of virtualization technology provides a way to solve this problem. Starting with the application scenario and definition of virtualization technology, this paper discusses the three most widely used VMM models and their corresponding virtualization schemes, and summarizes the advantages and disadvantages of these three different schemes on x86 architecture. Then it introduces the architecture and related technology of the mainstream paravirtualization virtual machine (Xen). Focusing on the memory management of Xen and combining with the source code, this paper studies the balloon driven algorithm for physical memory management, and analyzes the page jitter problem in the multi-virtual machine running, and aims at this problem. A centralized judgment balloon driving algorithm is proposed, the flow chart and pseudo code of the algorithm are given, and the performance of the virtual machine using the modified algorithm is tested and analyzed. Then, in order to evaluate an improved virtualization system objectively and reasonably and effectively, the author puts forward a "three-layer system performance evaluation model" which combines subjective and objective combination with AHP and deviation maximization method. The new versions of Xen Client2.1 and VMware Workstation of optimized memory management algorithm are used to evaluate the performance of these two desktop virtualization systems. Finally, the paper introduces the "centralized judgment balloon driving algorithm" and "three-layer system performance evaluation model" further improvement ideas and development prospects. The work of this paper can provide some reference for the research of memory management in Xen virtual machine system and the research of computer system performance evaluation, at the same time, it has certain reference value for the analysis and design of other systems.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333;TP302

【相似文献】