The Research Of Ceph Distributed File System Based On Object Storage

Because the traditional systems' interfaces use block or file interface.These models are difficult to meet the new requirements.Fortunately,Object-based storage systems come into being and have the block-based storage and file-based storage's advantages.Meanwhile,this system can storage and manage data automatically,and is being a hot spot research.Ceph is one of this kind of systems.As storage systems scale up,they consist of thousands of storage nodes.Multiple failures will occur frequently,which would induce a disaster because of data loss.How to design a efficient and reliable fault tolerance mechanism has become a issue that needs to be resolved urgently in distributed storage systems.The research of this paper is how to improve Ceph Distributed File System's reliability.The storage system mainly adopts redundancy storage strategy to ensure the system's reliability.It mainly use full backup and erasure codes storage method.Full backup strategy can be simply designed and support concurrent access,but this method costs multiple space consumption;erasure codes has strong fault tolerant ability and can save a lot of the space,but the computational cost and encoding and decoding will reduce the user access efficiency,this paper analyses the Ceph system's architecture and multi-replication redundancy method.On the basis of this storage strategy,we research the implement of erasure codes in Ceph system.According to the two tolerance mechanisms of the full backup and erasure codes,the system test environment is built,and the storage performance and safety performance of the two tolerance mechanisms are studied and compared.In order to increase the low repair rate of the default erasure code algorithm in the Ceph file system,the LRC erasure code algorithm is proposed in this paper.In the actual test,LRC erasure code algorithm performs perfect in improving the repair rate of the data and the average decoding rate.