Improving the Restore Performance via Physical-Locality Middleware for Backup Systems
Pengfei Li, Yu Hua, Qin Cao, Mingxuan Zhang · 2020
Data deduplication as an important middleware plays an essential role in current backup systems due to the high space efficiency, which however suffers from low restore performance, since the chunks are heavily fragmented. Existing systems leverage caching and rewriting schemes to reduce the number of container reads to restore the original data. However, when storing a large number of versions, the caching schemes become inefficient since each cached container consists of a few required chunks due to the exacerbated fragmentation, while the rewriting schemes consume more space to store the duplicate chunks.