refurbished cluster parts
Guide for refurbished cluster parts
Refurbished Cluster Parts List
Target: High CCU (100k+) Backend Compute on a Budget
Strategy: "Compute is cheap, State is hard". Offload DB to cloud, run stateless backend on cheap powerful refurb hardware.
1. Compute Nodes (The Workers)
Quantity: Start with 3, scale to 10+ as needed.
Why this hardware?
- Density: You can fit 10 of these in the space of a shoebox.
- Power: Idle at <10W each. Full load <35W.
- Cost/Perf: drastically cheaper than equivalent AWS EC2 vCPUs.
2. Networking
Goal: Low latency, high throughput for 100k connections.
3. Power & Accessories
4. Total Initial Investment (3 Node Cluster)
- Compute: 3 x $100 = $300
- Network/Misc: ~$55
- Total: ~$355
5. Deployment Role
- Node 1:
game-server-central+archival-worker(Master) - Node 2:
game-server-forest+game-server-mines - Node 3:
game-server-banks+promtail(Logging)