Proxmox VE

Proxmox VE is what sits between the metal and everything else on this page. It takes a server you own and turns it into a pool of virtual machines and containers, with backups, snapshots and live migration built in rather than sold separately.

Why we still choose it

Because virtualisation is the layer where licensing costs quietly become architecture. The commercial hypervisors meter by socket, by core, or by virtual machine, and the number changes at renewal — which is how a technical decision made in 2019 turns into a procurement emergency in 2026. Proxmox is AGPL, so the software does not expire and the bill does not scale with how successful you were.

The practical shape of it is a Debian system with KVM and LXC underneath, which matters more than it sounds. It means the host is an ordinary Linux machine you already know how to operate: standard packages, standard networking, standard logs. When something is wrong at three in the morning, you are debugging Linux rather than an appliance.

Clustering is included, not an upsell. Three nodes give you live migration between hosts, high availability for the guests that need it, and Ceph if you want the storage distributed as well. pvecm and qm do from the shell everything the web interface does, so the whole platform is scriptable and the runbook is a file rather than a sequence of screenshots.

Where it sits in our stack

Proxmox is usually the first thing installed and the last thing touched. NixOS guests run on top of it, which is a pleasant division of labour: Proxmox owns the hardware, the storage and the backup schedule, and each guest is still described entirely in one file. The firewall stays with pf at the border, and WireGuard carries the management network so the web interface is never on a public address.

Proxmox Backup Server is worth running alongside it. Incremental, deduplicated, verifiable backups with a retention policy that an auditor can read — and restores you have actually tested, because an untested backup is a rumour.

When it is the wrong answer

If you need one small always-on service, a single machine running it directly is simpler and we will say so. If your workload is genuinely elastic — spiky and unpredictable by an order of magnitude — a hyperscaler’s billing model may genuinely suit it better, and we would rather tell you that than sell you a cluster that idles.

And if nobody at your organisation wants to own hardware, that is a legitimate position. Proxmox on rented dedicated servers is the middle road we most often end up recommending: European providers, your own hypervisor, no per-socket licence, and no data centre of your own to staff.