Study Of Data Replication Technology In Cloud Storage Environment

Cloud storage is aimed at separating businesses and individuals from the heavy work of datastorage, provides the users simple and reliable storage service in the form of "pay on demand". Datareplication management is one of the key technologies which ensure the security and reliability ofdata in cloud storage. The research on replication management is an important aspect to improvethe quality of cloud storage.This thesis selects open source HDFS as cloud storage platform and analysis its features ofreplication management. To provide effective management of data replication on HDFS, this thesisproposes a new strategy determined the redundancy of data based on the hot of the data. The newstrategy firstly gives the method to calculate the hot of the file, and then changes the number ofcopies dynamically according to the value. At the same time the number of copies has themaximum value and the minimum value, the maximum value depends on the allowed costs ofconsistency maintenance, while the minimal value depends on the reliability requirements of theusers.Considering the huge differences between the cloud storage nodes, this thesis proposed a replicaplacement strategy based on TOPSIS to solve the decision problems with multi-objective. Thisstrategy uses rack perception on HDFS, takes comprehensive evaluation of multiple properties inmultiple nodes, calculates the distance between each alternative node to the best solution and theworst solution to deploy replica nodes, and can receive the high efficiency of replica nodes.