Research On Replication Strategy Of Distributed File System In Cloud Storage

With the rapid development of the Internet, data in the network growsexplosively，and the data storage has become a research hotspot in computer field.Because storage capacity and performance is limited, the applications of traditionalstorage methods, such as the SAN and NAS, are restricted. As taking the core ofdistributed file system, cloud storage has the advantage of cheap hardware and goodscalability. Replicas improve the reliability and availability of the system, but bring theproblem of load balancing and replica consistency.This thesis analyzes the typical distributed file systems combined with thebusiness feature of cloud storage, and studies the corresponding replica strategy whichfocuses on the efficiency and performance of the system accessing, load balancing andconsistency of data. The main work of the paper includes the following content:(1) The thesis proposes a new replica placement strategy based on consistencyhash algorithm. Replica location of traditional distributed file systems places mainlythrough path which stored in metadata server. When large amount of concurrent access,metadata server will become the bottleneck of the system.Replica placement strategybased on consistency hash algorithm is a good solution to the document retrieval,storage device extension and failure problem. The introduction of virtual nodemapping greatly reduces the data migration when system equipment has been changed.At the same time, increasing the weight attribute of equipment is a better solution tothe load balancing system.(2) The thesis proposes a new replica adjustment strategy based on file heat.Itdynamically adjusts the number of replicas based on the request of the files and theload of the server.And it enhances the performance and efficiency of the system. At thesame time, with the replica compression strategy, compressing the replicas which longtime have not been accessed, it ensures data reliability and availability, and savesstorage space of disks.(3) The thesis proposes a new replica consistency strategy based on user request.It maintains the consistency of replica and gives full consideration to the performance of the system. At the same time, in order to prevent that the inconsistent of long-termno user access documents makes the reliability of the data becoming lower, systemadopts the mechanism of regularly updated. When system is free, it will update thereplicas to a consistent state.