Free tool · 🧮 Proxmox / private cloud sizer
Size the cluster.Price it against the cloud.
The calculator our data-center practice uses for first-pass designs: workloads in, nodes out, with overcommit ratios, hypervisor reserve, replication overhead and N+1 — and an honest three-year TCO beside on-demand public cloud.
Rule of thumb built in: 8% RAM reserved for the hypervisor, 25% free space on storage, Ceph requires ≥3 nodes.
What it checks
Everything we would look at first on a client call.
Free, no sign-up, no email gate. If the report shows red, the fix is written next to it — and we are one click away if you would rather we did it.
- Node count from the binding constraint (CPU, RAM or storage) plus spare nodes
- Usable capacity after Ceph/ZFS/RAID overhead and free-space reserve
- Per-node CPU and RAM load so you can see the headroom
- Three-year TCOhardware, backup, network, subscriptions, power, colocation, operations
- Public-cloud equivalent at editable on-demand rates, with reserved-instance discount
- Payback period on the hardware
FAQ
Common questions
Why does Ceph need three nodes?
Ceph keeps three copies of every object and needs a quorum of monitors; with fewer than three nodes a single failure stops writes. For two-node sites use ZFS with replication and a QDevice for quorum.
What overcommit ratio is safe?
4:1 vCPU-to-thread is normal for web and application tiers; 2:1 for mixed production; 1:1 for databases and anything latency-sensitive. RAM is never overcommitted in this model.
Are the cloud prices accurate?
They are typical on-demand list prices for general-purpose instances in an Indian region, rounded. Put your own quote in — the point is to compare like with like, including the people cost on both sides.