Welcome to the Kreier Homelab

This is a documentation on my homelab projects. The setup currently consists of

• 3 routers,
• 1 managed switch,
• 3 always-on servers, and
• two LLM servers for ollama, n8n, openClaw and open WebUI.

Current Status

The lab features a combined 30 GB of GDDR5/X VRAM for local LLM inference and a dedicated Proxmox node for high-availability services.

Infrastructure at a Glance

LAN Network

• Raspberry Pi 3: Secondary network management.
• i5-8500T: The energy-efficient HP EliteDesk mini (4W idle).
• Jetson Nano 4GB: Accelerated workload with GPU and GbE network.

HOME Network

• Penta-GPU server: The LLM playground with quad NVIDIA GPUs.
• Raspberry Pi 4: Network backbone for Home Assistant and DNS.

Left: 2026-02 Quad-Nvidia GPU. Right: Stats from nvtop for the Penta-GPU server in the terminal.

Quick Links

• Setup Guide: Configuration and Docker basics.
• Penta-GPU server: Details on the 30GB VRAM LLM cluster.
• i5-8500: Proxmox and agentic AI. The energy-efficient Proxmox node.
• History: Benchmarks and the evolution of the lab.

---

Current Lab Architecture

Navigation Tip

In the sidebar to the left, click on Servers to expand the specific details for the Penta-GPU server and i5-8500T.

History

In early 2025 a prototype of the Penta-GPU server was running with a Gen1 x1 riser card and two weaker 6GB cards as a proof of concept. The LLMs were accessed in the terminal via ollama, no Open WebUI yet.

Left: The setup early 2025 with a GTX 960 on a riser card. Right: Ollama output and stats from nvtop in the terminal.