Research On Replica Management Strategy In Cloud Storage

Cloud computing profoundly changed the use of the services and resources, and cloudstorage as an important foundation for cloud computing, massive data storage and management,and to provide users with storage services. In the cloud storage system, usually use the replicatechnology to protect the high reliability, high availability and high scalability of the system. Byintroducing data replication, there are some problems such as replica created, the dynamicadjustment of replica and so on management strategies. Study in the cloud storage the replicamanagement strategy and validate its effectiveness has become an important research topic.The replica created problems include the number of replicas, replicas of position andcreation time, the dynamic adjustment includes migration, recreation, and deletion the problem.Reasonable replica management strategy needs to consider the storage cost and bandwidthconsumption and system load balancing, etc., before we could ensure data reliability, availability,and improve the quality of the cloud storage system performance, and cloud storage services. Areplica of the management strategy under the paper focuses on cloud storage environment, themain work of the following:1. Research the typical cloud storage systems, P2P and data grid and so on distributed filesystem management replica strategies, analyze the problems faced by a replica of themanagement in the cloud storage of systems engineering methods and ideas on the replicamanagement guidance role.2. Give a kind of in cloud storage node storing a replica of the evaluation method. Todifferences of the different needs and data node physical device conditions according to the userof data services, a reasonable evaluation of the data node, select nodes provide a reliable basisfor the replica placement.3. Based on a replica of the problem domain to create a strategy. Using a Hadoop-basedcloud storage system, divided according to the characteristics of the rack of the problem domain,to avoid the bottleneck problem of the name of the node; replica placed abstract multi-objectiveoptimization problem constraints, considering the replica placement optimization goals andconstraints to achieve the optimal replica distribution with a minimum price.4. Access based on user needs and files a replica of the heat and adjustment strategy. Themethod of solving a minimum number of copies based on user demand on the availability ofclear principles and procedures of the number of copies to add and delete. The policy can betterensure that the system load balancing and higher availability is verified by experiments.