Ceph relies on a layered architecture with a fast, distributed object store at its core. Any data written to Ceph over any of the supported protocols ultimately finds its way to the object store as a dispersed sequence of chunks. Data replication, data protection and redundancy are handled at this level.

 

The important concepts here are RADOS (Reliable Autonomic Distributed Object Store), which provides the underlying object store, CRUSH (Controlled Replication Under Scalable Hashing), a pseudo-random data distribution algorithm that works robustly in a heterogeneous cluster, and OSD (Object Storage Daemon), a massively parallel software component that operates at drive level, ensuring data encoding/decoding and internal data forwarding. Further concepts are discussed in this chapter.