The Research On Data Replication In Cloud Storage System

With the development of Cloud Computing and Cloud Storage technology,the volume of data in the system growth explosively.In the face of massive data storing and processing,Cloud Computing and Cloud Storage technology has become mainstream data share and processing technology in the current Internet technology.Data replication technology always be used for improving data availability,data fault tolerance and disaster recovery capability in Cloud storage systems.Key technologies of Cloud Computing include replication management technology and replica placement technology.These technologies not only bring benefits to the system but also bring some negative effects.The system stores multiple replicas increases bandwidth overhead.Replica maintenance also consumes some system resources.This thesis designs a data replication scheme and put forward a data replica placement based on load balance(RPBLB).(1)The data replication scheme concludes judging of replica creation time,replica number determination and replica consistency maintenance.The scheme uses the method based on block check to verify data and find the differences between the files need to be synchronized.In the process of data synchronization,only the different data need to be synchronized.(2)For replica placement problem,this thesis put forward RPBLB algorithm.It balances the whole data center's load and makes decision that if we need to create a new replica according to the distance between two data centers.If necessary,then take parameters of data center into account to calculate the capital value.Then choose the best data center to store the new replica.The simulation results show that RPBLB is feasible.Finally,this thesis designs some business scenarios based on the MapReduce and HDFS framework in the Cloud Computing platform Hadoop.It compares RPBLB with some other methods in same scenarios and the experiment results present that RPBLB is feasible.