云计算环境中虚拟机增量迁移技术的研究

发布时间：2018-12-21 20:59

【摘要】：作为近年来研究和应用的热点，云计算受到越来越多的IT企业和业内人士的关注，被认为是下一代互联网发展的新模式，将给公众的生产、生活带来巨大的变化。借助云计算，网络服务提供商可以短时间内处理千万甚至数亿计的数据，如同拥有超级计算机同样强大的处理能力。同时，使用者可以通过网络按需的使用这些资源和服务，实现将计算作为一种公共设施由公众使用的梦想，用户可以象使用水和电一样的使用云提供的服务。虚拟化是云计算系统的核心组成部分之一，它实现了IT资源的逻辑抽象和统一表示，，使得计算资源充分整合和高效利用，为云计算系统中资源的高效管理和动态分配分配提供了很好的解决办法。其中虚拟机实时迁移技术是最有价值的应用之一，也是本文的主要研究内容。虚拟机实时迁移是虚拟化技术中的关键技术之一。实时迁移是指在保持虚拟机运行的情况下将其从一台物理主机迁移到另一台物理主机上，该技术在负载均衡、系统容错、能耗管理等方面得到广泛的应用。迁移期间会有短暂的停机时间，用户不会受到明显的影响，但是在实时性要求比较高的环境中，如何降低迁移过程对用户带来的负面影响尤为重要。在本文中，作者选择开源的虚拟机监控器Xen作为研究平台，从如何减少迁移时间和停机时间的角度出发对Xen虚拟机迁移机制进行研究，并针对其不足进行了改进。本文的主要工作如下： 1）介绍了云计算和虚拟化技术的概念和发展状况，从服务器虚拟化、应用虚拟化等方面介绍了云计算和虚拟化技术之间密不可分的关系。 2）从虚拟机迁移的概念、迁移内容、迁移流程以及迁移评价指标等方面对实时迁移技术进行了详细的分析，并以Xen为研究对象，在深入研究了Xen架构、源码的基础上，指出其实时迁移机制在高脏页率的环境中的不足。 3）针对传统实时迁移框架的不足，本文提出了优化的实时迁移框架，主要是添加了两个模块：脏页统计模块和脏页分层模块。脏页统计模块的主要工作是：统计在预迁移过程中内存页面的变化情况，并以统计结果为标准判定工作集（Writable Working Set，WWS），解决了传统迁移机制在工作集选择上的片面性。脏页分层模块的主要工作是：根据脏页统计模块的统计结果，对内存页按照访问频率进行分层，对访问次数少的页面予优先，对访问次数多的页面延迟迁移，在一定程度上解决了迭代过程脏页被反复迁移的问题，有效的减少了迁移时间和停机时间。 4）在研究了Xen实时迁移源码的基础上，实现了脏页统计模块和脏页分层模块。分别提出了脏页统计算法和脏页分层算法，通过两者的协同工作实现缩短迁移时间和停机时间的目的。最后搭建实验平台，验证了算法的可行性和有效性，优化的实时迁移框架可以有效的改善在高脏页环境中的迁移性能。
[Abstract]:As a hot spot of research and application in recent years, cloud computing has attracted more and more attention from IT enterprises and industry insiders. Cloud computing is regarded as a new model for the development of the next generation Internet, which will bring great changes to the production and life of the public. With cloud computing, Internet service providers can process tens of millions or hundreds of millions of dollars of data in a short period of time, as powerful as supercomputers. At the same time, users can use these resources and services on demand through the network to realize the dream of using computing as a public utility. Users can use the services provided by the cloud as if they are using water and electricity. Virtualization is one of the core components of cloud computing systems. It realizes the logical abstraction and unified representation of IT resources and makes computing resources fully integrated and efficiently utilized. It provides a good solution for the efficient management and dynamic allocation of resources in cloud computing systems. Virtual machine real-time migration technology is one of the most valuable applications, and is also the main research content of this paper. Virtual machine real-time migration is one of the key technologies in virtualization technology. Real-time migration refers to moving virtual machine from one physical host to another while keeping it running. This technology has been widely used in load balancing, system fault tolerance, energy consumption management and so on. There will be a short downtime during the migration, and users will not be affected obviously. However, in the environment with high real-time requirements, how to reduce the negative impact of the migration process on users is particularly important. In this paper, the author chooses Xen, an open source virtual machine monitor, as the research platform to study the migration mechanism of Xen virtual machine from the point of view of how to reduce migration time and downtime. The main work of this paper is as follows: 1) this paper introduces the concept and development of cloud computing and virtualization technology, introduces the inseparable relationship between cloud computing and virtualization technology from the aspects of server virtualization and application virtualization. 2) from the concept of virtual machine migration, migration content, migration process and migration evaluation index, this paper analyzes the real-time migration technology in detail, and takes Xen as the research object, based on the in-depth study of Xen architecture, source code, etc. The shortcomings of its real-time migration mechanism in the environment with high dirty page rate are pointed out. 3) aiming at the shortcomings of traditional real-time migration framework, this paper proposes an optimized real-time migration framework, which mainly includes two modules: dirty page statistics module and dirty page delamination module. The main work of the dirty page statistics module is to calculate the changes of the memory pages in the process of pre-migration, and to judge the work set (Writable Working Set,WWS) according to the statistical results, which solves the one-sidedness of the traditional migration mechanism in the selection of the working set. The main work of the dirty page delamination module is as follows: according to the statistical results of the dirty page statistics module, the memory pages are stratified according to the access frequency, the pages with less visits are given priority, and the pages with more visits are delayed to migrate. To some extent, it solves the problem that dirty pages are repeatedly migrated during iteration, and effectively reduces the migration time and downtime. 4) on the basis of studying the source code of Xen real time migration, the dirty page statistics module and dirty page delamination module are implemented. The dirty page statistics algorithm and the dirty page stratification algorithm are put forward, which can shorten the migration time and the downtime by the cooperation of the two algorithms. Finally, an experimental platform is built to verify the feasibility and effectiveness of the algorithm. The optimized real-time migration framework can effectively improve the migration performance in high dirty pages environment.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP302

【参考文献】