Private Cloud
We build a cloud on OpenNebula and Ceph over dedicated servers: VMs, volumes and S3 at hardware cost rather than at a hyperscaler's list price.
Included
- OpenNebula on KVM: virtual machines, networks, quotas and self-service for your team through the web UI and the API
- Ceph under VM disks (RBD) and object storage (RGW, S3): three nodes or more, with no single point of failure
- Servers at Hetzner, OVH, Selectel or on hardware you already own — ordered on your account, not ours
- A private network between nodes and sites, with every control panel off the public internet
- Cluster and disk monitoring that alerts on Ceph degradation before it turns into data loss
- VM and bucket backups to a second site, with restore drills
- A hybrid with AWS where it earns its keep: a VPN between the two, and workloads moved piece by piece
How the work runs
We start with the bill and the load: what runs today, where, what it costs, and where it hits a ceiling. If you do not need a cloud and three servers with Docker would do, you hear that at this step, not after the build.
Then the design: how many nodes, which disks, how many copies Ceph keeps, where the second site is. We pick servers at a provider to fit the budget and order them on your account — the cloud stays yours even if we stop working together.
We build the cluster with our own automation, test a node failure and a disk failure before any workload moves, then migrate VMs and data in parts, in agreed windows.
After that we operate it: OpenNebula and Ceph upgrades, disk replacements, growing the cluster, monitoring, and scheduled restores.
If you already run Ceph
You can start with a fixed-price cluster check: health, replica count, placement rules, disks and versions — ending in where the risk of losing data sits. Remote access is optional: an export you take yourself from our list of commands is enough.
What we need from you
Access to the current infrastructure or the cloud bill, a list of what has to move, and one person who decides how much downtime is acceptable. The account at the server provider is yours; we help choose the hardware and place the order.
Where this does not apply
If you have a dozen VMs living quietly at one provider, a cloud platform adds more work than it saves. Ceph pays off from three nodes and at a volume worth running distributed storage for; below that, plain disks with replication and backups are enough. We will say so.
What this draws on
These areas of the catalogue are covered by the package, or alongside it.