"Syrup" - an Ubuntu Server
"Syrup" - an Ubuntu Server
Syrup is my personal Ubuntu-based Linux server lab — built to function as a stable, always-on infrastructure platform for experimentation, automation, media services, and local AI workloads.
It began as a compact Mini-ITX system focused on GPU acceleration and local LLM experimentation. Over time, it evolved into a full ATX-based infrastructure platform designed for simplicity, performance, and long-term reliability.
Rather than treating it as a one-off build, Syrup has become an evolving system — each revision solving architectural frustrations from the previous one.
#===================================================================================================================================#
Mark 1 – Compact, GPU-Focused, Experimental
Syrup Mk1 was a Mini-ITX Ubuntu server built primarily around:
• Local AI workloads (RTX 3060 12GB)
• GPU experimentation
• Always-on containerized services
It ran reliably, but storage architecture and container workflows were influenced by external NAS behavior and Synology constraints. Over time, I found myself fighting abstractions rather than working directly with Linux.
Mk1 was the proof of concept, Mk2 became the refinement.
#==================================================================================================================================#
Syrup Mk2 represents a deliberate architectural reset.
I moved away from appliance-style NAS behavior and layered RAID configurations (mdadm + btrfs stacks) in favor of a simpler, more transparent Linux-native approach.
Hardware was upgraded to a full ATX platform:
Ryzen 7 7700X
32GB RAM
RTX 3060 12GB
2× 2TB SATA SSDs (hot storage)
4× 8TB HDDs (cold storage)
The goal was clarity and control.
Mk2 runs a straightforward Ubuntu Server base with Docker-managed services including:
Jellyfin (media streaming)
Home Assistant (automation)
Glance dashboard
Glances monitoring
Valheim server
SMB file shares
Obsidian vault hosting
Hoarder bookmarks
Despite the workload variety, the system operates with low CPU utilization and significant headroom, functioning as a stable, quiet infrastructure backbone.