Research On Backup Service Quality For Multi-user Deduplication

According to IDC,global data storage demand in 2020 will reach 44 ZB.Today,every minute of every day,204 million emails are published,4 million searches are generated by Google,and 2.46 million Bits of content are shared by Facebook.As a result,companies need to continually purchase large amounts of storage to meet growing storage needs.However,due to the existence of some problems such as equipment procurement budget,data management cost,energy consumption and the space occupied,enterprises can't simply solve the problem by improving storage capacity.The industry has therefore proposed deduplication technology,which is widely used in backup systems.In multi-user backup system,users with different preferences and needs can enjoy different system services due to different preferences and needs.Therefore,users are divided into different levels,namely hierarchical services.However,there are shared data blocks between different users,and the shared data block between multiple users only needs one user to read and write.How to uniformity amortize the overhead of reading and writing shared data blocks to each user to ensure the hierarchical service of the system is this paper need to solve.In the study of backup service quality under multi-user deduplication.The existing high-performance level service quality research mainly guarantee hierarchical services in data backup.Specifically includes two aspects: System resource adjustment and shared data processing.However,this solution can't adjust system resources in real time,at the same time,the overhead of processing shared data blocks between users can't be uniformity amortization to each user.In other words,the solution can't guarantee the hierarchical service of the system effectively.Aiming at the shortcomings of existing hierarchical service schemes,this paper proposes a shared data block processing scheme based on data classification in multi-user data backup which is called DCSDP.The scheme uses different processing method for different types of shared data blocks,which can more accurately amortization shared data block processing overhead to each user.It also can reduce overhead of fingerprint retrieval and improves system backup performance.In multi-user data recovery,data fragmentation prediction technology based on sliding window and high level priority processing strategy are proposed in this paper.On the one hand,it can guarantee hierarchical service when multi-user recovery,on the other hand,it can reduce the impact of data block fragmentation on recovery performance to some extent.The experimental results show that the shared data block processing scheme based on data classification improves the overall backup performance by 26.63% compared with the existing high-performance hierarchical service quality scheme,and the hierarchical service of the system is more obvious.Relative service errors between users of same level decreased by 3.71%,which can better guarantee the quality of service between users of same level.In the case of multi-user data recovery.The overall performance of the system increased by 16.6% compared to the lack of shared block processing and data fragmentation processing,and hierarchy service of the system becomes more apparent.Relative service errors between users of same level decreased by 4.99%,which can better guarantee the quality of service between users of the same level.