Research And Implementation Of Cloud Storage System Based On Moose File System

Cloud storage is a derivative from cloud computing, with all features of cloudcomputing. With cloud computing emerging, cloud storage which is a part of cloudcomputing develops and gradually becomes a research hotspot. Load balancing is a keytechnology of cloud computing, so merging load balancing into cloud storage issignificant.MooseFS(Moose File System), a fault-tolerant, scalable, POSIX(PortableOperating System Interface of Unix)-compatible, multipurpose distributed file system.In addition, MooseFS is C language implementation of GFS(Google File System). AsMooseFS owns these features, MooseFS is chosen as the file system of cloud storage.This thesis completes the following work based on GCI Science and TechnologyCO., Ltd:1. MooseFS’s latest source code(1.6.26) is read carefully, the problem of existingload balancing algorithm on MooseFS is analyzed and based on which a novel loadbalancing algorithm is proposed, named Performance-Load(PL) dynamic load balancingalgorithm. PL algorithm is compared with the existing algorithm, including efficiencyof algorithm and experimental results. For high performance, PL algorithm is improved.2. According to the actual project, cloud storage system based on MooseFS iscompleted. MooseFS cloud storage system is analyzed and is devided into two parts.The first is storage subsystem and the second is monitoring subsystem. In storagesubsystem, building scheme of MooseFS is designed and high availability for metadataserver is implemented. Deployment automation module to build MooseFS automaticallyis developed and finally MooseFS read-write performance is tested. In monitoringsubsystem, framework of monitoring subsystem is analyzed and designed, storagecontrol function module is designed and implemented, and the storage control module istested.