CephFS

CephFS is a distributed POSIX-compatible file system built on the Ceph storage platform. It provides scalable, fault-tolerant shared access to large data sets by separating metadata management from data storage. File data is stored in Ceph as objects within RADOS pools, while directory trees, file permissions, and other metadata are managed by a dedicated cluster of metadata servers.

Architecture: CephFS uses three main components. The data plane stores user data as objects in RADOS and is accessed by clients through the RADOS data path. The metadata plane consists of one or more metadata servers (MDS) that hold the namespace and metadata for files and directories. The client interface, available as a kernel client or FUSE module, talks to MDS for metadata and to OSDs for data. The overall namespace can be scaled by adding MDS daemons and by distributing metadata across them; data scaling is achieved automatically by RADOS and CRUSH.

Key features include snapshots, quotas, and POSIX semantics. CephFS supports multi-MDS deployments for scalability and high availability, with active-active metadata servers and failover. Data integrity is provided by RADOS replication or erasure coding, while MDSs coordinate metadata operations to preserve consistency. Clients can mount CephFS clusters across nodes, enabling shared access for workloads such as HPC, analytics, and cloud storage.

Administration involves monitoring Monitors, OSDs, and MDS daemons, tuning network and storage pools, and using Ceph's management tools and dashboards. CephFS remains part of the Ceph project and is maintained through open-source releases, with ongoing improvements in performance, reliability, and scalability.