Ceph unifies object, block, and file storage in a single cluster. One system covers S3-compatible object storage, block devices for virtual machines, and a shared filesystem, and it scales from a few terabytes to hundreds of petabytes without changing architecture.
Where our experience comes from
We have run Ceph in production since the Argonaut release โ the first stable version, long before most of the tooling around it existed. Ansgar Jazdzewski, who runs opHUB, is a recognized Ceph Ambassador.
That time has been spent across the whole stack rather than one corner of it. Block storage for virtualization via RBD, iSCSI, and NVMe. Shared filesystems through CephFS with NFS and SMB gateways in front. S3 object storage at multi-tenant scale, which is where the interesting problems live: bucket index behavior, storage class placement, and what happens when a bucket holds millions of objects.
When stock Ceph does not do what a cluster needs, we adjust it. That includes patching radosgw-admin where an operation was missing an option we needed for a fleet-wide migration.
What’s included
- Design and deployment of Ceph clusters, from terabytes to 100+ PB
- Cluster sizing, capacity planning, and hardware guidance (HDD/SSD/NVMe)
- RBD, iSCSI, and NVMe storage for virtualization and Kubernetes
- CephFS with NFS/SMB for shared file services
- S3 / RGW object storage at scale, including multi-tenancy
- Performance tuning, Bluestore optimization, and custom integrations
- Monitoring, failure testing, and ongoing operations support
How we work
Every cluster gets a documented design before hardware is ordered, and a runbook your team can operate from afterward. We test failure paths deliberately โ pulling disks, losing nodes, filling pools โ because a storage cluster nobody has seen fail is a storage cluster nobody knows how to recover.