The hardware, services, and methodology behind my homelab architecture.
Published
1 Servers
Hostname
Model
CPUs (total cores)
Memory
GPUs
PSUs
edge-router
Dell PowerEdge R410
2 × Xeon L5520 @ 2.26 GHz (8 cores total)
16 GB DDR3 ECC
–
2 × 500 W
docker-ai
Dell PowerEdge R720
2 × Xeon E5‑2670 v2 @ 2.5 GHz (40 cores total)
256 GB DDR3 ECC
2 × NVIDIA Tesla P40 24 GB
2 × 750 W
docker-logic
Dell PowerEdge R630 (8 SFF)
2 × Xeon E5‑2680 v3 @ 2.4 GHz (48 cores total)
128 GB DDR4 ECC
–
2 × 750 W
Both servers run Ubuntu 24 LTS with a single, root‑based Docker daemon. The edge router runs pfSense and also hosts a WireGuard tunnel for remote/public access to the internal network.
2 AI Server (docker-ai) Containers
Container
Image
GPU
Host Ports
Notes
Portainer
portainer/portainer-ce:lts
–
9443 (HTTPS)
Runs as root (-u 0:0) to avoid cert‑dir permissions
Ollama
ollama/ollama:latest
✅
11434
Models stored in /var/containers/ollama (bind‑mount)
Stable‑Diffusion WebUI
zefie/stable-diffusion-automatic1111:latest
✅
7860
Model assets bind‑mounted from /var/containers/sd
3 Logic Server (docker-logic) Containers
Container
Image
Host Ports
Notes
Portainer
portainer/portainer-ce:lts
9443 / 8000
Runs as root inside
OpenWebUI
ghcr.io/open-webui/open-webui:latest
3000
OLLAMA_BASE_URL=http://docker-ai:11434
Dashy
lissy93/dashy:latest
8088
Config & icons bind‑mounted from /home/cj/docker/dashy
Jellyfin
jellyfin/jellyfin:latest
8096
Config & cache in /srv/jellyfin; media on local disks