Phase 2

Hardware

Choosing your hardware based on what you plan to self-host is the first critical step.

This guide covers quite a bit of ground, but at the same time, it only really scratches the surface - there are a ton of self-hostable services out there that you can run, so don’t feel like you have to do everything here all at once.

For example, if you just want to start out with setting up something like:

  • Navidrome music server to replace Spotify
  • FileBrowser for storing a few gigabytes of files to replace Google Drive
  • Paperless-ngx for backing up important documents

Then you could technically get away with something as small and cheap as a Raspberry Pi with 4GB RAM and a 128GB microSD card, though it wouldn’t necessarily be my first choice.

If you want to do a bit more with your server in addition to the above, like:

  • Streaming your movies and TV libraries with Jellyfin
  • Automate your music/movies/TV management with Sonarr/Radarr/Lidarr
  • Automatically back up photos and videos from your phone using Immich
  • Running a Minecraft server for you and your friends using Crafty

Then you would probably want to step up to something like this mini PC with a 4-core CPU and 16GB RAM, or maybe even this mini PC with an 8-core CPU and 24GB RAM.

If you want to run everything detailed in this guide and more, taking advantage of room for multiple hard drives for lots of storage capacity, and even adding a dedicated GPU to run local AI models offline, then you will likely want to build your own server or repurpose an old gaming PC.

Below is a quick-and-dirty separation of performance tiers based on workload:

Low CPU/RAM:

  • NGINX web servers (for building and hosting your own websites)
  • FileBrowser (Google Drive replacement)
  • Paperless-ngx (document backup & organization)
  • Navidrome (Spotify replacement)

Decent CPU + RAM:

  • Crafty (running your own Minecraft server)
  • n8n (workflow automation)

Decent CPU with GPU/iGPU + plenty of storage:

  • Jellyfin (media streaming - Intel CPU + iGPU with QuickSync OR AMD CPU + separate NVIDIA GPU with NVENC recommended)
  • Sonarr/Radarr/Lidarr (media library automation)
  • Immich (Google Photos replacement)

Fast CPU + Dedicated GPU with high VRAM + fast SSD:

  • Ollama + OpenWebUI (running local AI models)
A) Starter (experimenting with a few services):
  • CPU: 4–8 cores
  • RAM: 8GB
  • Storage: 500GB – 1TB SSD
B) Media + Automation:
  • CPU: 6–12 cores (Intel CPU + iGPU with QuickSync OR AMD CPU + separate NVIDIA GPU with NVENC)
  • RAM: 16GB (32GB if you can)
  • Storage: 500GB – 1TB NVMe + HDDs for media
C) The “Everything” Build:
  • CPU: 8–16 cores
  • GPU: NVIDIA RTX w/ high VRAM
  • RAM: 32GB (64GB if you can)
  • Storage: 1–2 TB NVMe (for VMs + containers + databases) + 2–8 HDDs for ZFS pool

Storage Strategy

If you decide to build a server with multiple HDDs for large storage capacity, plan for redundancy:

  • 2 drives → mirror
  • 3–5 drives → RAIDZ1
  • 6+ drives → RAIDZ2

Avoid SMR hard drives. Use CMR hard drives only.

“RAIDZ? SMR? CMR? ZFS?”

Don’t worry - we will get into this later when we get to TrueNAS.

Single Machine vs Separate Storage & Compute

For this guide, we will only be considering the scenario of a single machine. We assume a desktop or old gaming PC so you can use:

  • Multiple internal hard drives
  • An NVMe SSD
  • Optionally a GPU

BUT you could technically build two separate machines if you wanted:

  • Machine A: Bare-metal TrueNAS SCALE with multiple HDDs
  • Machine B: Mini PC running Proxmox + Debian for compute

Pros: easier upgrades, redundancy and separation
Cons: more hardware and power, more complex to set up

Rackmount

If you want hot-swap hard drive bays and actual enterprise server components, used rackmount servers can be found on eBay for attractive prices, but expect more noise and higher power consumption.