《基于影子缓存的多增量虚拟机启动系统 (1)》由会员分享，可在线阅读，更多相关《基于影子缓存的多增量虚拟机启动系统 (1)（52页珍藏版）》请在金锄头文库上搜索。
1、A Thesis Submitted in Partial Fulfillment of the Requirementsfor the Degree of Master of EngineeringA Starting System ofMultiple Incremental Virtual MachinesBased on Shadow CacheCandidate： Xiukun HuMajor： Computer Software and TheorySupervisor： Associate Professor Xia XieHuazhong University of Scien
2、ce and TechnologyWuhan 430074, P.R.ChinaMar, 2013独创性声明本人声明所呈交的学位论文是我个人在导师指导下进行的研究工作及取得的研究成果。尽我所知，除文中已经标明引用的内容外，本论文不包含任何其他个人或集体已经发表或撰写过的研究成果。对本文的研究做出贡献的个人和集体，均已在文中以明确方式标明。本人完全意识到，本声明的法律结果由本人承担。学位论文作者签名：日期：年月日学位论文版权使用授权书本学位论文作者完全了解学校有关保留、使用学位论文的规定，即：学校有权保留并向国家有关部门或机构送交论文的复印件和电子版，允许论文被查阅和借阅。本人授权华中科技大学可
3、以将本学位论文的全部或部分内容编入有关数据库进行检索，可以采用影印、缩印或扫描等复制手段保存和汇编本学位论文。本论文属于保密，在_年解密后适用本授权书。不保密。（请在以上方框内打“”）学位论文作者签名：指导教师签名：日期：年月日日期：年月日华中科技大学硕士学位论文摘要在虚拟机镜像文件集中存储的环境下，大量虚拟机同时启动时，由于硬盘 I/O吞吐能力有限，使得虚拟机镜像文件从硬盘传输到内存的时间过长，最终导致虚拟机启动速度过慢。为了解决上述问题，基于影子缓存的多增量虚拟机启动系统 Scache（StartingSystem of Multiple Incremental Virtual Machi
4、nes Based on Shadow Cache）利用虚拟机间的相似性，使用基础镜像文件加增量镜像文件的方式为虚拟机提供虚拟块设备，然后通过搭建由多个数据集缓冲结构组成的影子缓存，以及修改开源虚拟机管理系统 Xen 4.1.2上 Qemu块设备模拟器中增量镜像文件异步读机制，将增量虚拟机启动过程中需要的数据单独缓存于特定的数据集缓冲结构中以供增量虚拟机启动时取用。其中，每一个数据集缓冲结构对应一种基础镜像文件，可被基于此基础镜像而创建的多个增量虚拟机共用。由此对于虚拟机而言，内存的缓存结构被设计得和硬盘中镜像文件的层次结构一致，犹如镜像文件栈的影子。与此同时， Scache系统还避免了多个增
5、量虚拟机启动过程中被共用的基础镜像文件里部分数据被重复读取的情形。经测试验证，Scache系统可最小化增量虚拟机启动过程中硬盘和网络上的负载，并对启动过程有较明显的加速效果。而针对一种基础镜像创建的多个增量虚拟机，Scache系统额外消耗的内存在具体数量上也很小，且不受虚拟机内存配置大小的影响。关键词：虚拟化，增量镜像文件，缓存，启动过程I华中科技大学硕士学位论文AbstractWhen a large number of virtual machines are started at the same time in theenvironment where virtual disk ima
6、ges are stored centrally, the time will be long which ittakes for virtual disk images to be transferred from the hard disk to the memory becausethe hard disk I/O throughput capacity is limited. This eventually leads the virtual machinestart-up speed to be very slow.In order to solve the problem ment
7、ioned above, the starting system of multipleincremental virtual machines based on shadow cache, also named Scache, exploits thesimilarity between virtual machines used by consumers to provide virtual machines withvirtual block devices which are composed of basic image files and incremental image fil
8、es.And through modification of asynchronous reading operations in Qemu block devicemodel on Xen 4.1.2 and building shadow cache that consist of multiple data-set-cachestructures, this system stores the data needed by the starting process of incremental virtualmachines in a specific data-set-cache st
9、ructure so as to get incremental virtual machines toshare it during their starting process. Here every data-set-cache structure in the shadowcache corresponds to one of the basic images and is shared by multiple incremental virtualmachines that are created based on this basic image. Under such circu
10、mstances for virtualmachines, the memory cache hierarchy structure and the hard disk image file stackstructure is a one-to-one relationship, as if the former is the latters shadow. MeanwhileScache system also avoids repeatable read operations on the same data in shared basicimage files during starti
11、ng process of incremental virtual machines.By concrete evaluation experiments, Scache system is able to minimize the load onhard disks and network during the starting process of incremental virtual machines.Moreover the starting process of incremental virtual machines has been acceleratedobviously i
12、n the role of this system and regardless of the configured memory size, thissystem costs only a small amount of extra memory on a basic image file shared by severalincremental virtual machines.Keywords: virtualization, incremental image files, cache, starting processII华中科技大学硕士学位论文目录摘要. IABSTRACT . II1绪论1.1研究背景 . (1)1.2国内外研究现状 . (3)1.3研究内容 . (7)1.4论文的组织结构 . (9)2基于影子缓存的多增量虚拟机启动系统总体设计2.1系统概述 . (10)2.2设计思想 .
