Research On Optimization Of Space Efficiency And Restore Performance For Data Backup System

Deduplication has been widely used in data backup system to eliminate redundant data,which speeds up the backup process in the system and reduces storage overhead.In order to preserve the locality of data streams,deduplication packs multiple chunks into a large,fixed-size container as a storage unit in the system.However,a large number of containers are written to disks without being filled due to a large amount of metadata generated by small files.Unfilled containers lead to more containers needed to store a backup,which decreases storage efficiency and exacerbates restore performance.In order to improve restore performance,existing schemes selectively rewrite some fragmented chunks,but a large number of repeated calculations during the rewriting process lead to a decrease of rewriting speeds.To solve the problem of low storage efficiency,a metadata-region adaptive container structure,called MACS,is proposed.MACS maintains a tag to record the length of metadata area in the container.The boundary between metadata area and data area is dynamically divided to ensure the maximum space efficiency of the container.In order to solve the problem that MACS is not applicable for existing deduplication systems,the container management and replacement strategy has been re-implemented.In order to efficiently identify data fragments,a fast sliding window-based rewrite approach,called FSW,is proposed.FSW adopts metadata chunks to record the container reference count to avoid repeated traversal of chunks,thus reducing the impact of rewriting on the speed of deduplication.Experimental results demonstrate that the rewriting speed of FSW is higher than the existing rewriting methods based on sliding windows.The container loading ratio of MACS is significantly higher than that of the traditional container structure.Moreover,MACS achieves better restore performance than traditional container structure.When used in combination with FSW,MACS can achieve a good trade-off between deduplication ratio and restore performance.