← All talks

Session X · 15 April 2026

Proxmox walkthrough

Presented by Pablo Pan

An open-source virtualisation platform for fast, cheap test environments, and a plan to put it on idle office hardware.

What was covered

  • What it is. An open-source virtual environment. Alongside regular virtual machines it offers LXC containers, lightweight and handy for spinning up a machine quickly to test something.
  • What you need. A CPU with virtualisation support, at least 16 GB of RAM in practice (the official minimum is lower), and a wired network connection. Wi-Fi is not suited to it. Installation takes about ten minutes from an ISO on a USB drive.
  • The vocabulary. Nodes (the hosts), virtual machines, Linux containers, storage, pools (like resource groups, useful for permissions) and clusters. A cluster needs at least three hosts, so Pablo advised against one at home.
  • Virtual machine or container? A VM is fully isolated, which suits heavier workloads. A container shares the host kernel and suits simple apps such as nginx or a Python service. Containers need privileges to run, which carries a security risk.
  • Deploying quickly. Cloud-init images give you preconfigured VMs within minutes. Pablo uses Terraform to create resources from templates, and the Linux bridge connects everything.
  • What it is good for. Training, self-hosted GitHub Actions runners, firewalls and home services, with no dependence on outside providers. A live demo created a container in seconds, and the only limit is your own hardware.

What came up

  • Community scripts automate many deployments, and Docker can run inside a virtual machine. Proxmox works as an isolation layer: try something, and delete the machine if you do not like it.
  • Idle office hardware. The guild owns a machine that is no longer used for projects. The proposal was to install Proxmox on it to test large models, give team members their own machines (interns included) and keep a fallback for when cloud tokens run out.
  • GPUs are tricky. Passing a GPU through needs a virtual machine, and only one GPU per machine on that hardware. Installation would be manual, and the next step was to speak to the machine’s owner.

Also on the day

A hands-on AI workshop with the Google Developer Group in Porto on 28 April was announced, using Google ADK for beginners. The documentation and slides were to be shared with the guild.