Research On Data Replication Strategy In Cloud Storage

Cloud storage is based on the distributed file system,grid computing,cluster applications,and data storage services is provided through the application of software integration on a large number of different types of storage devices in network.In order to enhance the reliability and fault tolerance of the system,the replica technique is adopted in the cloud storage.However,the existing replica management strategy is further studied in the aspects of system reliability and economic cost.On the existing replica management,either only one target is focused on considering,and other targets is regarded as constraints,or the single target of all targets is tried to be synthesized by means of linear weighting method which has the disadvantages that the size of the weights is needed to know in advance,the weight is relied on experience,the different scenarios are needed to solve many times,and the solution of the overall situation is hard to receive when multiple optimization objectives are simultaneously taken into account.In view of the above problems,the following works are carried out in this paper.Firstly,the research status of the replica management strategies are described,and their shortcomings are analyzed.The number of copies of the static replica management strategy are relatively fixed.And the number of copies of the dynamic replica management strategy is dynamically adjusted,but the cost and reliability of the system are not considered on the copies of add and delete operations,at the same time,dynamic replica management strategy is generally only aimed at a specific application scenarios.Either the property of the system is only optimized,or the relationship is neglected among the multiple objectives which can be transformed into a single objective when dealing with multiple optimization objectives when copies are placed.And the research contents are introduced in this paper.Secondly,a multi-objective optimization of data replica management strategy(MRM)is proposed.And the new ideas are presented to solve the copy management on the concept of Pareto.Through the detailed analysis of the mutual relations among the common objectives,two obvious conflict objectives which are system reliability and storage costs are selected.Aiming at the problem of multiple conflicting objectives,MRM is designed and implemented to search for the relations between the number of copies and the layout of the copies on NSGA-II.Because NSGA-II is based on real number coding,and the coding in this paper is discrete,so coding method,operator and other details of the adaptive improvement are presented with MRM on NSGA-II.New parameters and variables(such as weights)are not introduced,and the optimal solution is gained no need to depend on experience,and the system is not affected by determining the weight that is not accurate based on MRM algorithm.Meanwhile,the optimal solution set is only needed to obtain in a run,and the suitable optimal solution is selected on the basis of different.Finally,MRM strategy is effectively verified through the experiments.According to different application scenarios,the most appropriate number of copies of the current system characteristics and layout strategy is chosen in the Pareto optimal solution set.Besides,its adaptability is wider,and user's personalized needs are better satisfied.