An Effective RAID Data Layout for Object-Based De-duplication Backup System
An Effective RAID Data Layout for Object-Based De-duplication Backup System
复制标题
基于对象的重复数据删除备份系统的有效RAID数据布局
DOI：
10.1049/cje.2016.06.021
复制
发表时间：
2016
影响因子：
1.2
通讯作者：
Zheng Jun
中科院分区：
文献类型：
--
作者：
Yan Fang;Tan Yu'an;Zhang Quanxin;Wu Fei;Cheng Zijing;Zheng Jun
本文提出了一种用于基于对象的重复数据删除备份系统的非对称分组数据组织（AGDO）方法。基于对象的重复数据删除是检测复合文件重复数据的有效解决方案。设计了一种非对称分组策略，将阵列中的磁盘划分为不同的组，在每个组中采用并行的数据访问策略，不同类型的对象存储在不同的组中，组大小可能不同，通过动态磁盘组调整算法可以设置和管理动态组大小。AGDO的性能进行了评估，并证明是足够的连续存储应用程序。其结果是，磁盘访问在很长一段时间内集中在一部分磁盘上，并在10个磁盘配置中将功耗降低到25%。此外，基于对象的重复数据消除与AGDO相结合，在提高复合文件的数据恢复速度方面具有很大的潜力。我们已经证明，这种组合使平均恢复速度提高了11%。
We propose an alternative Redundant array of independent disks (RAID) data layout, Asymmetrical grouping data organization (AGDO), for object-based data de-duplication backup system. Object-based data deduplication is an effective solution for detecting duplicate data for compound files. We designed an asymmetrical grouping strategy, the disk in the array are partitioned into different groups, and in each group parallelism data access scheme is adopted, different types of objects are stored in different groups with maybe different group size, it can set and manage dynamic group size by using dynamic disk group adjustment algorithms. The performance of AGDO is evaluated and proved to be sufficient for the continuous storage application. The result is that disk accesses are concentrated in a part of the disks over a long time period and reduces the power consumption to 25% in a 10 disk configuration. Moreover, object-based de-duplication combined with AGDO has great potential in increasing data restoration speed for compound files. We have shown that this combination makes average restoration speed improved 11%.