U4 Compute

Choosing a platform: hypervisor, NAS OS or plain Linux

The first operating system you install on a homelab machine shapes everything that follows. There is no single right answer, but there is a right answer for what you want to do. This guide compares the four common foundations and explains how to choose.

At a glance

  • Choose by goal. Learning virtualization, running a few services simply and storing data safely each point to a different foundation.
  • Plain Linux with a container engine is the smallest thing to learn. A hypervisor platform adds snapshots and room to experiment.
  • Keep the hypervisor host dedicated to virtualization. Do not install applications on it.
  • Check a project's health, release dates and license before you build on it.
  • Keep data separate from the platform, and changing your mind later stays cheap.

The four common foundations

FoundationWhat it isStrengthsTrade-offs
Open-source hypervisor platformRuns virtual machines and lightweight system containers side by side, managed from a web interfaceSnapshots, scheduled backups, ZFS integration, clustering once you have three machinesTwo layers to understand; more to configure before your first service runs
ZFS-based NAS distributionA storage-first operating system that serves files and can run some applications on the sideExcellent data integrity, snapshots and replication, friendly sharing setupApplications are secondary; needs at least 8 GB of memory and two matching drives to be useful
NAS and container appliance OSA storage operating system that pools mixed drives and offers an application catalogVery easy to start, mixed drive sizes, one-click applicationsUsually licensed by drive count; less flexible for complex designs
Plain server Linux with a container engineA minimal server distribution plus containers defined in compose filesSmallest mental model, most tutorials, maximum flexibility, transferable skillsNo built-in snapshots or web interface unless you add them

A fifth option deserves a mention: self-hosting layers that add an application store and dashboard on top of an ordinary Linux install. They can be pleasant, but they add another dependency between you and your services. At least one popular layer stopped releasing updates in 2025, which left its users to migrate. If you use one, make sure your data and compose files would survive without it.

Choosing by goal

If your main goal is toStart with
Learn virtualization, try many operating systems and break things safelyA hypervisor platform
Run a handful of services for the household with the least to learnPlain server Linux with containers
Store family files and photos safely, with some apps on the sideA ZFS-based NAS distribution on its own machine
Combine mixed drives and apps on one box with minimal effortA NAS and container appliance OS
Prepare for a job that uses a particular platformThat platform, or the closest open equivalent

Two good ways to begin

Our editors disagree on this, which tells you both answers are reasonable.

Plain Linux first

  • One layer to learn, so your energy goes into containers, not the platform.
  • Everything you learn applies to every Linux server you will ever touch.
  • Move to a hypervisor later, when you want several virtual machines.
  • Walkthrough: Your First Self-Hosting Weekend.

Hypervisor first

  • Snapshots make every experiment reversible in seconds.
  • Try other operating systems without new hardware.
  • Run your containers inside one dedicated VM and keep the host clean.
  • Walkthrough: Mini PC Hypervisor Build.

Virtual machine, system container or application container

A hypervisor platform gives you three ways to run a service. Choose per service, not once for the whole lab.

  • Virtual machine when the service needs its own kernel or a different operating system, ships as a virtual appliance, needs strong isolation, or you are testing system-level changes. Home automation hubs often run here, as do firewall VMs for learning; the household firewall belongs on its own device.
  • System container when the service is a light Linux workload that you manage like a small server, and you want fast startup and low memory use.
  • Application container when the project publishes an official image, you want repeatable deployments from a compose file and you may move it to another machine later.

A common and sensible pattern is one dedicated Linux VM running the container engine, holding most application containers, with a few VMs and system containers alongside it for services that need them. See Containers and Virtualization.

Habits that keep any platform healthy

  • Keep the host dedicated. Do not install applications directly on a hypervisor host. Maintenance, upgrades and recovery stay simple when the host only does virtualization.
  • One service per container or VM. Updating one application never breaks another, each can be snapshotted on its own and a runaway service only consumes its own allocation.
  • Do not hand out all the memory. Start small containers at 512 MB to 1 GB and raise limits only when monitoring shows real pressure. The host and its storage need headroom.
  • Use a separate administrative account for routine work rather than the account created during installation.
  • Never expose the platform's management interface to the internet. See Keep Management Off the Internet.

Check a platform's health before you commit

  • Release dates and support windows. Major versions reach end of life. A previous major release of one popular hypervisor platform went out of support in 2026, and tutorials written for it still rank highly in searches. Check that a tutorial matches the version you install.
  • Activity. Recent releases, active forums and responsive maintainers.
  • License and funding. Is it open source? Is there a paid tier, and what happens if the free tier changes?
  • Appliance support. Some commercial virtual appliances are only supported on particular hypervisors. Their disk images can usually be converted for another platform, but you may lose vendor support.
  • Community install scripts. Collections of one-line scripts that create containers and VMs on a hypervisor save time, but they run as root on your most important machine. Read each script before running it and write down what it installed.

Storage on a single machine

If one machine must provide both virtual machines and file shares, avoid nesting storage systems inside each other while you are learning. Running a NAS operating system inside a VM requires passing the physical disks or their controller through to it, and mistakes there are how data gets lost. A simpler design lets the hypervisor manage the disks with ZFS directly and runs a small file-server container on top. Better still, give storage its own machine once you can. See Storage and NAS and Shared Storage for Containers.

Changing your mind later

You will probably change platforms at least once, and that is fine. Keep application data on separate storage, keep compose files and configuration in git, and keep good backups. Then moving to a new platform is an afternoon of reinstalling and reattaching data, not a rebuild from memory. See Automation and Configuration.