Ceph Capacity
Sizer.
Estimate RADOS capacity for host-level CRUSH failure domains. Results are planning estimates, not a substitute for testing or a Ceph design review.
Topology matters
in Ceph.
By default, the sizer reserves one host’s share of theoretical pool capacity. This models a uniform cluster that can absorb one host failure without filling the remaining hosts beyond their planned capacity. With three equal hosts and replicated size 2, that reduces 46.1 TB theoretical capacity to 30.7 TB before any additional reserve or growth.
This model assumes host-level CRUSH failure domains and evenly balanced OSD capacity. It does not calculate PG count, recovery time, latency, device endurance, compression, BlueStore DB/WAL sizing, or custom CRUSH rules.
Ceph documentation recommends recovery headroom, at least 10 Gb/s networking, and planning RAM for BlueStore OSDs plus operating-system and recovery margin. Validate the final layout with the actual CRUSH map and workload tests.
Read Ceph hardware recommendations