Zhou and Colleagues Greedily Improve the Cloud

Data deduplication has been broadly used in the Cloud due to its storage space saving ability.  One issue with deduplication is a phenomenon called data fragmentation.  To deal with data fragmentation, Cloud implements a procedure that diminishes the restore performance.  Although capping methods have been developed to alleviate data fragmentation, they employ rewriting procedures that are only partially successful.  To address this problem, TSYS School assistant professor of computer science Yi Zhou and his research colleagues from Jinan University (China) and the University of Exeter (United Kingdom) propose a multi-segment greedy rewriting method named MGRM.  As they explain in their study forthcoming in IEEE Transactions on Cloud Computing, MGRM works sequentially to rewrite in a way that achieves a good balance between deduplication and restore performance.  To do so, it adaptively switches between two working modes – an optimal rewriting mode and a radical rewriting mode – according to workload.  The optimal rewriting mode utilized by MGRM improves deduplication, while the radical rewriting mode is utilized to improve restore performance.  Thus, unlike existing capping methods that improve restore performance at the cost of deduplication, MGRM pays attention to both aspects.  Extensive experimental results presented in the study show that MGRM achieves high restore performance, coupled with high deduplication.  In particular, compared to two state-of-art schemes, MGRM improves deduplication and restore performance by up to 114.83% and 99.34%, respectively.