mirror of
https://github.com/altstackHQ/altstack-data.git
synced 2026-04-17 21:53:12 +02:00
146 lines
4.7 KiB
Plaintext
146 lines
4.7 KiB
Plaintext
---
|
||
title: "Hardware & VPS Sizing"
|
||
description: "How much RAM, CPU, and disk you actually need for self-hosting. VPS provider comparison and scaling strategies."
|
||
---
|
||
|
||
# Hardware & VPS Sizing
|
||
|
||
The #1 question new self-hosters ask: **"What server do I need?"**
|
||
|
||
Short answer: less than you think to start, more than you think once you're hooked.
|
||
|
||
## Quick Sizing Guide
|
||
|
||
### How Much RAM Do I Need?
|
||
|
||
| Setup | RAM | What You Can Run |
|
||
|---|---|---|
|
||
| **Starter** | 2 GB | 1–2 lightweight tools (Uptime Kuma, Plausible) |
|
||
| **Hobbyist** | 4 GB | 3–5 tools + a database + reverse proxy |
|
||
| **Power User** | 8 GB | 8–12 tools + multiple databases |
|
||
| **Homelab** | 16 GB | Everything + AI models (small ones) |
|
||
| **AI Workloads** | 32+ GB | LLMs, image generation, video AI |
|
||
|
||
> 💡 **Start with 4 GB.** You can always upgrade. Most VPS providers let you resize without downtime.
|
||
|
||
### CPU Guidelines
|
||
|
||
| Workload | vCPUs Needed |
|
||
|---|---|
|
||
| Static tools (Uptime Kuma, PocketBase) | 1 vCPU |
|
||
| Web apps (Plausible, Outline, n8n) | 2 vCPUs |
|
||
| Heavy apps (PostHog, Supabase, Metabase) | 4 vCPUs |
|
||
| AI inference (Ollama, Stable Diffusion) | 4+ vCPUs + GPU |
|
||
|
||
### Disk Space
|
||
|
||
| Component | Typical Usage |
|
||
|---|---|
|
||
| Base OS + Docker | 5–8 GB |
|
||
| Each Docker image | 100 MB – 2 GB |
|
||
| PostgreSQL database (small app) | 500 MB – 5 GB |
|
||
| Log files (unmanaged) | 1–10 GB |
|
||
| AI models (per model) | 4–70 GB |
|
||
|
||
**Minimum recommended:** 50 GB SSD.
|
||
**Comfortable:** 80–160 GB SSD.
|
||
**AI workloads:** 200+ GB NVMe.
|
||
|
||
## VPS Provider Comparison
|
||
|
||
| Provider | Starting At | Pros | Best For |
|
||
|---|---|---|---|
|
||
| [**DigitalOcean**](https://m.do.co/c/2ed27757a361) | $6/mo (1 GB) | Simple UI, great docs, predictable pricing | Beginners |
|
||
| **Hetzner** | €3.79/mo (2 GB) | Best price-to-performance in EU | Power users, EU hosting |
|
||
| **Contabo** | €5.99/mo (4 GB) | Cheapest for RAM-heavy setups | Budget homelab |
|
||
| **Linode (Akamai)** | $5/mo (1 GB) | Reliable, good network | Small projects |
|
||
| **Vultr** | $5/mo (1 GB) | Global locations, hourly billing | Testing and experimentation |
|
||
| **Oracle Cloud** | Free (4 vCPUs, 24 GB ARM) | Unbeatable free tier | Zero-budget hosting |
|
||
| **Home Server** | One-time cost | Full control, unlimited bandwidth | Privacy maximalists |
|
||
|
||
> 🏆 **Our Pick:** [DigitalOcean](https://m.do.co/c/2ed27757a361) for beginners (simple, reliable, [$200 free credit](https://m.do.co/c/2ed27757a361)). **Hetzner** for best value. **Oracle Cloud free tier** if you want to pay nothing.
|
||
|
||
## Real-World Stack Sizing
|
||
|
||
Here's what actual AltStack setups typically need:
|
||
|
||
### The Bootstrapper Stack (4 GB RAM)
|
||
- Coolify (deployment platform)
|
||
- Plausible (analytics)
|
||
- Uptime Kuma (monitoring)
|
||
- Listmonk (newsletters)
|
||
- Caddy (reverse proxy)
|
||
|
||
### The Privacy Stack (4 GB RAM)
|
||
- Vaultwarden (passwords)
|
||
- Jitsi Meet (video calls)
|
||
- Mattermost (messaging)
|
||
- Caddy (reverse proxy)
|
||
|
||
### The AI Stack (16–32 GB RAM)
|
||
- Ollama (LLM inference)
|
||
- Stable Diffusion (image generation)
|
||
- TabbyML (code completion)
|
||
- Continue.dev (AI coding)
|
||
|
||
## Scaling Strategies
|
||
|
||
### Vertical Scaling (Bigger Server)
|
||
|
||
The simplest approach. Just resize your VPS:
|
||
|
||
- **DigitalOcean:** Resize droplet (takes ~1 minute)
|
||
- **Hetzner:** Rescale server (may require reboot)
|
||
- **Home server:** Add RAM sticks
|
||
|
||
### Horizontal Scaling (More Servers)
|
||
|
||
When one server isn't enough:
|
||
|
||
```
|
||
Server 1: Databases (Postgres, Redis)
|
||
Server 2: Application containers
|
||
Server 3: AI workloads (GPU)
|
||
```
|
||
|
||
Connect them with a private network (most VPS providers offer this for free) or a VPN like WireGuard.
|
||
|
||
### The "Start Small" Strategy
|
||
|
||
1. **Month 1:** $6/mo droplet (1 GB) — Deploy 1–2 tools
|
||
2. **Month 3:** Resize to $12/mo (2 GB) — Add more tools
|
||
3. **Month 6:** Resize to $24/mo (4 GB) — Running your full stack
|
||
4. **Month 12+:** Add a second server or move to Hetzner for better value
|
||
|
||
## Monitoring Your Resources
|
||
|
||
Always know how much headroom you have:
|
||
|
||
```bash
|
||
# Quick resource check
|
||
free -h # RAM usage
|
||
df -h # Disk usage
|
||
nproc # CPU cores
|
||
uptime # Load average
|
||
|
||
# Docker resource usage
|
||
docker stats # Live container metrics
|
||
docker system df # Docker disk usage
|
||
```
|
||
|
||
## Red Flags
|
||
|
||
🚩 **RAM constantly above 90%** → Resize or move a service to another server.
|
||
|
||
🚩 **Disk above 80%** → Clean Docker images (`docker system prune -f`) or resize disk.
|
||
|
||
🚩 **CPU at 100% for extended periods** → Check which container is the culprit with `docker stats`.
|
||
|
||
🚩 **Swap usage above 1 GB** → You need more RAM. Swap is a band-aid, not a solution.
|
||
|
||
## Next Steps
|
||
|
||
→ [Quick Start](/quick-start) — Deploy your first tool
|
||
→ [Deploy Guides](/deploy) — Browse 65+ tools
|
||
→ [Docker in 10 Minutes](/concepts/docker-basics) — Foundation knowledge
|