Skip to content
← All Posts
Homelab

Building a 24-Service Homelab on a $400 NAS

How I replaced $300/month in cloud subscriptions with a TerraMaster F2-424 running Docker — and why the process taught me more than any SaaS product ever will.

The $300/Month Wake-Up Call

Last year I sat down and added up what I was paying for cloud services. Notion for notes. Otter.ai for transcription. Todoist for tasks. Bitwarden cloud for passwords. Google Photos for storage. A handful of automation tools. RSS readers. Monitoring dashboards.

The total came to roughly $300 a month. For a single user. For tools that held my data on someone else's servers, behind someone else's terms of service, subject to someone else's pricing decisions.

I decided to build the alternative.

The Hardware Decision

I didn't start with a rack server or a custom PC build. I bought a TerraMaster F2-424 — a compact 2-bay NAS with an Intel N95 processor and 16GB of RAM. Total cost: around $400 with drives.

This isn't a monster machine. It's a small box that sits on a shelf and draws about 15 watts idle. But with TrueNAS Scale and Docker, it runs 24 services simultaneously — and has for months without a hiccup.

Hardware Specs

  • Device: TerraMaster F2-424
  • CPU: Intel N95 (4 cores, up to 3.4GHz)
  • RAM: 16GB DDR5
  • Storage: 2x NVMe (apps) + 2x HDD (bulk storage)
  • Power draw: ~15W idle, ~35W under load
  • Monthly electricity: ~$3-4

The Stack: 24 Services, One Box

Everything runs as Docker containers managed through Docker Compose. Here's what's running:

Document & Knowledge Management

  • Paperless-ngx — OCR-powered document management. Every receipt, letter, and manual gets scanned, tagged, and made searchable.
  • BookStack — Self-hosted wiki for structured documentation. Runbooks, procedures, reference material.
  • Linkwarden — Bookmark manager with full-page archiving. Never lose a useful link again.
  • FreshRSS — RSS reader. Curated feeds instead of algorithmic noise.

Automation & Orchestration

  • n8n — Visual workflow automation. 80+ workflows handling everything from document processing to morning briefings.
  • ntfy — Push notifications. Every important event across the stack sends a notification.

AI & Vector Search

  • Qdrant — Vector database for semantic search. Embeddings generated on my Windows PC, stored and queried on the NAS.
  • AnythingLLM — RAG system. Upload documents, ask questions, get cited answers.

Media & Photos

  • Plex — Media server for movies and TV.
  • Immich — Self-hosted Google Photos replacement with face detection and smart albums.

Security & Infrastructure

  • Vaultwarden — Bitwarden-compatible password manager.
  • Uptime Kuma — Service monitoring with alerting.
  • Syncthing — File sync between devices (Obsidian vault, shared folders).
  • Actual Budget — Privacy-first budgeting tool.

Plus a handful of supporting services: Cloudflare Tunnels for secure external access, AdGuard Home for DNS-level ad blocking, a homelab dashboard, and more.

What I Learned

RAM Is the Real Bottleneck

With 16GB total, memory management matters. Immich alone can consume 3GB during photo processing. I learned to stagger heavy operations — run backups at night, limit concurrent transcription jobs, and stop AnythingLLM when I'm not actively using it. A simple free -h command became part of my daily routine.

Docker Compose Is the Best Documentation

Every service is defined in a Docker Compose file. That file IS the documentation — it captures every port mapping, volume mount, environment variable, and dependency. When something breaks, I read the compose file, not a wiki article. When I want to rebuild, I run one command.

Automation Compounds

The first n8n workflow I built saved me 5 minutes a day. The tenth saved me an hour. The thirtieth replaced an entire category of manual work. Automation isn't linear — each workflow you build makes the next one easier because you've already solved the shared problems (LLM calls, notification routing, error handling).

Privacy Is a Feature, Not a Sacrifice

Self-hosting isn't about paranoia. It's about ownership. My documents aren't training someone else's AI model. My passwords aren't in a breach notification. My photos aren't being scraped for ad targeting. And when a cloud service shuts down or changes pricing, I don't lose anything.

The Cost Comparison

Monthly Costs: Cloud vs. Self-Hosted

Category Cloud Self-Hosted
Notes & Knowledge $15/mo $0
Transcription $30/mo $0
Password Manager $5/mo $0
Photo Storage $10/mo $0
Automation Tools $50/mo $0
AI APIs $40/mo $0
Misc SaaS $50/mo $0
Electricity $4/mo
Total ~$200/mo ~$4/mo

Hardware cost (~$400) pays for itself in 2 months.

Should You Do This?

If you're a developer or IT professional, yes — without hesitation. The learning alone is worth it. You'll understand Docker, networking, Linux administration, API design, and system architecture at a level that no tutorial can teach.

If you're not technical, it's still possible — tools like TrueNAS, CasaOS, and Umbrel make it increasingly approachable. Start with one service (Paperless-ngx or Vaultwarden are great entry points) and grow from there.

The homelab isn't just a cost savings exercise. It's a statement: your data, your infrastructure, your rules.