Study On Quality Of Backup Service Based On The Multi-Tenant Data Deduplication

In the traditional backup systems based on sing-user deduplication,it only deletes the redundant data emerging in the backup streams under the same user,without considering the redundant data across multiple users.While in the cloud backup systems for multiuser data deduplication,it is necessary not only to delete redundant data within a single user,but also to delete redundant data among multiple users.Such deletion of the redundant data leads to large amounts of data shared among multiple users.At the same time,a large number of users uses the cloud backup systems,and the quality of the cloud backup service required by the user is diversified.However,the duplicate data shared by the multiple users seriously affects the quality of the backup service requested by the user,which is mainly reflected in two aspects:(1)Uneven distribution of resources.During the data backup process,the processing of the user's backup data streams consume many resources,such as memory and disk.However,each user's backup data stream has different data redundancy.According to the principle of data deduplication,when the amount of data and the user level are the same,the users with higher data redundancy have less demand for resources,while the users with low data redundancy need more resources.Under the condition of limited system resources,if each user's resources are allocated according to the total amount of data chunks in the backup data stream,the users with high data redundancy would have excessive supply of resources,while users with low data redundancy suffer from resource deficiencies.(2)The quality of backup service level cannot be guaranteed.Data deduplication among multiple users causes a large amount of duplicate data to be shared between users.When multiple users backup/restore the shared data,and uses the principle of first-comefirst-served as in the traditional backup system,it would probably result that the highlevel users store/read these shared data,while the Low-level users do not need to process them.As a result,the backup performance of low-level users would be higher than that of the high-level users,making it impossible for users to obtain a fair level of backup service quality.To solve these problems,this paper proposes a high-performance hierarchical quality-of-service guarantee strategy based on data deduplication,which is called HPLQoS.HPLQoS includes two aspects,a hierarchical resource allocation mechanism and a hierarchical shared duplicate data processing mechanism.Hierarchical resource allocation mechanism,which dynamically adjusts the hierarchical resource parameters among users according to the user's real-time service quality and the use of resources to meet user-level resource requirements,have solved the problem of the excessive supply of resources due to the data deduplication.Hierarchical shared duplicate data processing mechanism is mainly to monitor the duplicated data emerged between multiple users,determine and decide which duplicate data should be stored/read by which user within a predetermined period to ensure the fair level of backup service quality across multiple users.The experimental results show that the deviation of the quality of service guarantee level of HPLQoS is about 5%,which can meet the user's level of backup service quality requirements.