关键词: 对象文件系统, Ceph, 数据容错机制, 纠删码, 完全副本

Abstract:

This paper focuses on the reliability and security of the Ceph distributed file system. The reliability and security of the system are realized through the data fault tolerance mechanism. The widely used data fault tolerance techniques include full copy scheme and erasure code scheme. Full copy scheme copies the data into multiple copies for distributed storage, which reduces the storage efficiency and wastes the storage space. Erasure codes scheme can use the storage space efficiently, but it need store the data encoded, and decode the data when accesses them, which makes the data access efficiency low. Combining the two data fault tolerance mechanisms, on the basis of the Ceph distributed file system, this paper designs a data placement strategy based on the file level. The strategy both provides the full copy scheme and the erasure code scheme, which users can select the appropriate storage scheme based on the file level. The strategy ensures that the important data have high security and can effectively use the storage space.

Key words: object-based file system, Ceph, data fault tolerance mechanism, erasure code, full copy