Notice
This document is for a development version of Ceph.
Architecture
These pages explain how Ceph stores data, keeps it available when hardware fails, and serves it to clients.
How Ceph Stores Data
The Ceph Storage Cluster: RADOS, the object store under every Ceph service.
Scalability and High Availability: CRUSH, the cluster map, Monitor quorum, and authentication.
Dynamic Cluster Management: pools, placement groups, peering, rebalancing, and scrubbing.
Erasure Coding: how chunks are written and read.
How Clients Use It
Ceph Clients: object, block, and file interfaces.
Ceph Protocol: librados, watch/notify, and striping.
Extending Ceph and Legacy Features
Ceph Object Classes: extending the OSD with your own object methods.
Cache Tiering (Deprecated): kept for clusters that still run it.
Why This Matters for Sizing
Heartbeats, peering, rebalancing, and recovery run on the OSD hosts, so every server needs CPU, RAM, and network for them. See Hardware Recommendations.
Next Steps
Additional Resources
Brought to you by the Ceph Foundation
The Ceph Documentation is a community resource funded and hosted by the non-profit Ceph Foundation. If you would like to support this and our other efforts, please consider joining now.