MMSD: a Metadata-Aware Multi-Tiered Source Deduplication Cloud Backup System in the Personal Computing Environment
Haiyan Meng, Jing Li, Weiqing Liu, Changchun Zhang · International Review on Computers and Software (IRECOS) · 2013
The deduplication efficiency/overhead ratio of existing source deduplication solutions in cloud backup systems, in despite of their complex deduplication workflows, was not ideal due to their insufficient study of file semantics. In this paper, we present MMSD, a metadata-aware multi-tiered source deduplication cloud backup system in personal computing environment, to obtain an optimal tradeoff between the deduplication efficiency and deduplication metadata storage overhead and finally achieve a shorter backup window than existing approaches. MMSD makes full use of file metadata including file size, type, timestamp, path information, modification frequency, and the advantage of whole file level deduplication compared with chunk level deduplication, and efficiently combines these two deduplication levels. Our experimental results with real world datasets show that, compared with the state-of-art source deduplication methods, MMSD can improve the overall deduplication efficiency from 75% to 91% with only 33.8% of deduplication metadata storage overhead and shorten the backup window by at least 54.2%.