虚拟机备份系统中存储空间的优化

发布时间：2018-07-18 08:07

【摘要】：随着科学技术的快速发展，越来越多的企业或单位开始利用虚拟机搭建数据中心来应对日益增长的计算需求。然而，随着数据中心虚拟机数目的不断增加，单个虚拟机发生故障从而导致整个数据中心失效的概率也随之增加。目前，虚拟机保护技术主要包括虚拟机迁移和虚拟机备份，它们都是迁移或备份完整的虚拟磁盘文件，这会消耗大量的存储资源和网络带宽。因此，提高存储资源的利用率，是虚拟机保护技术中必须解决的一个重要问题。重复数据删除技术是一种新型的数据缩减技术，它能够消除相同数据内容的文件、数据块或字节，只保存它们的一个副本。因此，采用基于数据段相似性的重复数据删除解决方案消除虚拟磁盘文件之间的重复数据，不仅可以降低虚拟磁盘文件的存储空间，而且能够减少虚拟机的备份窗口。针对重复数据删除技术的读性能下降和指纹匹配问题，，分别采用VmStore和HashDB技术。VmStore遵循相邻数据相邻存储的原则，可以有效缓解重复数据删除所带来的读性能问题。HashDB基于数据段的相似性原理，不仅可以提高指纹匹配效率，而且能够节约内存空间。对包含有多种虚拟磁盘文件的数据集进行了详细测试。测试结果表明，变长分块算法适用于不同虚拟平台下的虚拟磁盘文件，同一虚拟平台下的不同虚拟磁盘文件之间宜采用固定长分块算法，基于数据段相似性的重复数据删除解决方案可以有效地挖掘出重复数据，并且通过采用HashDB和VmStore技术极大地改善了指纹匹配问题和读性能问题。
[Abstract]:With the rapid development of science and technology, more and more enterprises or units have begun to use virtual machines to build data centers to cope with increasing computing needs. However, with the increasing number of virtual machines in the data center, the probability of the failure of a single virtual machine and the failure of the whole data center is also increased. Computer protection technology mainly includes virtual machine migration and virtual machine backup, which all migrate or backup complete virtual disk files, which will consume a lot of storage resources and network bandwidth. Therefore, improving the utilization of storage resources is an important problem that must be solved in the virtual machine protection technology.
Duplicated data deleting is a new technique for data reduction, which can eliminate files, blocks, or bytes of the same data content and only save a copy of them. Therefore, duplication of data deletion solutions based on data segment similarity can be used to eliminate the duplication of data between virtual disks and not only reduce virtual magnetism. The storage space of the disk file and the backup window of the virtual machine can be reduced. In view of the decline of read performance and fingerprint matching in repeated data deleting technology, the principle that VmStore and HashDB technology.VmStore follow adjacent data adjacent storage can be used to effectively alleviate the read performance problem of.HashDB based on the number of repeats. According to the principle of similarity, the efficiency of fingerprint matching can not only be improved, but also the memory space can be saved.
The test results show that the variable length block algorithm is suitable for the virtual disk files under different virtual platforms, and the fixed long block algorithm should be used between the different virtual disk files under the same virtual platform, and the repeating data deleting solution based on the similarity of the data segment is used. The case can effectively extract duplicate data, and by using HashDB and VmStore technology, fingerprint matching and reading performance problems are greatly improved.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP309.3;TP302

【参考文献】