Design And Implementation Of Ceph-based Enterprise Cloud Storage Systems

In recent years,with the rapid development of cloud computing and the Internet of things,traditional enterprise storage(such as SAN and NAS)has been unable to cope with the storage demand of large scale unstructured data.Thus,object-based storage has set off a new wave of development after network access storage,block-based storage and file-based storage.Through analyzing the present situation and development trend of enterprise storage,this thesis studies how to design and implement a set of object storage system with high availability and high scalability based on the current popular open-source SDS(Software-Defined Storage),i.e.,Ceph distributed storage technology.The objective is to satisfy the storage and management requirements of enterprise to the unstructured data such as pictures and videos in the private cloud,public cloud and hybrid cloud scenarios.Firstly,in the preliminary technology study,the advantages of object storage were clarified by comparing the principles and characteristics of traditional network storage and object storage.This thesis also investigates the typical solutions proposed by the SDS community,such as Sheepdog,GlusterFS and Swift.Most importantly,by depthly studing on the Ceph technology architecture and RadosGW object gateway,this thesis clarifies the technical selection,and establishes the interface and communication mode of the system through the overall analysis of Amazon S3(Simple Storage Service)protocol.In the system design stage,this thesis extracts the functional use cases and non-functional requirements based on the usage scenarios of storage services for enterprise applications,deal with the preliminary architecture and outline design accordingly,and abstracts the two subsystems of account management service and S3 storage service.In the stage of system deployment,this thesis mainly focuses on the development of Ceph custom package,minimizes the extraction and packaging on the object storage component and the dependant files after compiling the source code,and developes a complete deployment operation and maintenance tool to minimize the rapid installation of the cluster.In the development stage,this thesis gradually implements the SDK development and application through the S3 API debugging and encapsulation of entity classes.After the system deployment and the client joint debugging,this thesis also performs the performance evaluation and stability evaluation on the storage system.The evaluation results demonstrates that this system can meet the expected non-functional requirements on the basic performance and reliability.Finally,in view of the increasing new requirements of the enterprise cloud storage,this thesis summarizes and shares the optimization scheme in the actual production environment and the best practices for extending the operation and maintenance capability,and the service capability,which has certain practical reference value for enterprise users who use the Ceph system.