Managed Ceph Storage Clusters
We design, install and operate Ceph storage clusters in our privately owned Cambridge Tier III data centre — petabyte-scale object, block and file storage with no single point of failure, on hardware we own and a network we run.
PB-scale
Per cluster
3x / EC
Replication or erasure coding
S3 / RBD / FS
One cluster, three APIs
24/7
UK NOC operations
Software-defined storage, properly engineered
One self-healing cluster that grows with you — block, object and file on the same fabric, operated by our UK NOC.
Ceph at the Core
We design, install and operate Ceph clusters end-to-end — RADOS, RBD, CephFS and S3-compatible RGW on hardware we own and rack in Cambridge.
No Single Point of Failure
Self-healing, self-balancing object placement. Lose a disk, a node or a whole rack and the cluster keeps serving reads and writes.
Petabyte-Scale Capacity
High-density JBODs with enterprise HDDs for cold/warm tiers and NVMe OSDs for hot tiers — start at tens of terabytes, scale to petabytes.
NVMe Performance Tier
Dedicated all-NVMe Ceph pools for databases, VM images and latency-sensitive workloads. Hundreds of thousands of IOPS, sub-millisecond reads.
S3-Compatible Object Storage
Ceph RGW gives you a private S3 endpoint inside the UK — drop-in for backups, media, ML datasets, Veeam, Restic, rclone and more.
Block & File Too
RBD block devices for Proxmox, KVM and Kubernetes (CSI), plus POSIX CephFS for shared file workloads — one cluster, three interfaces.
Replication & Erasure Coding
Choose 3x replication for hot data or erasure coding (e.g. 8+3) for capacity-efficient archives. Per-pool, tuned to your workload.
UK Data Sovereignty
Every byte stays in our privately owned Cambridge Tier III data centre on our AS48825 network — no hyperscaler, no overseas replicas.
One cluster, every storage interface
Ceph speaks block, object and file natively. We expose the right interface to each consumer — no separate silos to manage.
| Interface | Typical consumers | Notes |
|---|---|---|
| Ceph RBD (Block) | Proxmox, KVM, OpenStack, Kubernetes CSI | Thin-provisioned, snapshots, clones, live-migration friendly |
| Ceph RGW (Object / S3) | Backups, media, ML/AI datasets, archives | S3 & Swift APIs, bucket policies, versioning, lifecycle |
| CephFS (File) | Shared POSIX filesystems, render farms, HPC scratch | Multi-MDS, snapshots, quotas, kernel & FUSE clients |
| iSCSI Gateway | VMware, Windows Server, legacy SAN consumers | Multipath, ALUA, on top of RBD |
| NFS Ganesha | Linux/Unix NFS clients on top of CephFS or RGW | NFSv3 / NFSv4, HA-deployable |
What our customers store on Ceph
From private-cloud back-ends to petabyte media archives — same cluster pattern, sized to the workload.
Private cloud storage
Back your Proxmox, OpenStack or Kubernetes platform with one resilient Ceph cluster instead of a fragile SAN — block, object and file on the same fabric.
Backup & archive targets
An on-net, UK-sovereign S3 endpoint for Veeam, Restic, rclone, Duplicacy and Bacula — with immutability and lifecycle rules.
AI / ML datasets
High-throughput Ceph pools to feed GPU training jobs, with object storage for raw datasets and CephFS scratch for active runs.
Media & broadcast
Capacity-efficient erasure-coded pools for video archives, with an NVMe tier for active edit and ingest workflows.
How we build a Ceph cluster
Every cluster is engineered to your durability, performance and growth targets — not pulled off a template. We run current upstream Ceph (Reef / Squid) on hardware we own, with a dedicated cluster network and proper CRUSH topology from day one.
- Right media for the tier — NVMe OSDs for hot pools and DB/WAL, enterprise HDDs for capacity, NVMe metadata for CephFS.
- Dual networks — separate public and cluster (replication) networks on dual 10/25/100 Gbps for predictable recovery performance.
- Failure-domain aware CRUSH — placement across hosts, racks and (optionally) rooms so the cluster survives real-world failures, not just disk loss.
- Monitored 24/7 — PG health, OSD latency, scrub status, near-full thresholds and version drift — handled by our NOC, not your team.
Example reference cluster
Topology : 6 x OSD nodes, 3 x MON/MGR, 2 x RGW, 2 x MDS
Per OSD node: 2 x AMD EPYC, 256 GB RAM
24 x 18 TB enterprise HDD (capacity tier)
4 x 7.68 TB NVMe (DB/WAL + hot tier)
Network : 2 x 25 GbE public, 2 x 25 GbE cluster (LACP)
Raw / usable: ~2.6 PB raw, ~860 TB usable (3x replica)
~1.7 PB usable (EC 8+3)
Software : Ceph Reef/Squid, RBD + RGW (S3) + CephFS
Encryption : LUKS at-rest on every OSD, TLS on RGW
Management : Fast2host UK NOC, 24/7 monitoring & opsConfigurations are illustrative — every cluster is sized to your capacity, IOPS, durability and growth targets.
How to get started
01
Sizing & design
Tell us usable capacity, IOPS/throughput targets, durability goals and growth plan. We design replica vs erasure-coded pools, CRUSH topology, MON/MGR/MDS layout and the right OSD media mix.
02
Build & install
We rack the hardware in Cambridge, cable a dedicated cluster network, install and harden Ceph (Reef/Squid), bootstrap the cluster and configure RBD, RGW or CephFS to fit your stack.
03
Operate & scale
Ongoing monitoring, OSD replacement, version upgrades and capacity expansion — handled by our UK NOC. Scale by adding nodes; the cluster rebalances itself.
8 common questions