Ceph Storage Architecture for OpenStack

Context

Cloud storage architecture

Role

Storage engineering

Challenge

Balance capacity, performance, and failure-domain requirements across OpenStack environments with different infrastructure constraints.

50+ TB distributed storage · 3 / 5 / 6-node architectures

• Ceph
• RBD
• Cinder
• SSD / NVMe

Context

The OpenStack environments used more than 50 TB of distributed Ceph storage. Infrastructure requirements varied between locations, so the storage architecture was not reduced to a single fixed topology.

Architecture considerations

The work covered three-node, five-node, and six-node Ceph architectures depending on the environment. Design decisions included OSD and MON placement, failure-domain thinking, and the operational effect of node count and hardware layout.

Separate SSD and NVMe pools supported storage-tier and workload-separation requirements. Ceph RBD integrated with OpenStack Cinder through volume types, with frontend QoS used where workload controls were required.

Engineering focus

• Match storage topology to infrastructure and failure-domain constraints.
• Separate performance-oriented and capacity-oriented workloads through pool design.
• Keep Ceph architecture aligned with OpenStack volume types and operational workflows.
• Treat monitoring, maintenance, and recovery paths as part of the design.

No unsupported benchmark or performance figures are presented here; the published scale and topologies are the verified details.