High-Performance Builds / In Progress

Home Compute Cluster

A small self-hosted cluster running containers, data jobs and lab services — the personal 'supercomputer' behind the experiments.

Home server rack with blue status lights in a dark room

Three nodes, shared storage and a segmented network. It runs the data pipelines, dashboards and automation experiments that would otherwise cost real money in the cloud.

It is also the best teaching tool I own: nothing forces you to understand a system like being the person who has to bring it back up.

Why build it instead of renting it

Cloud is the right answer for production. For learning, a bill that grows every time you experiment quietly teaches you to experiment less.

Owning the hardware moves the cost to the front and makes curiosity free after that.

What it actually runs

The workload mix is deliberately close to the professional work it supports.

Scheduled ETL jobs feeding a local warehouse.
Dashboard and BI services for testing data models.
Isolated security lab segment for authorised testing only.
Model and automation experiments that need a GPU on tap.

Specification

Nodes: 3 × compute nodes, mixed generation
Total cores: 40 physical cores across the cluster
Memory: 256 GB aggregate ECC
Storage: ZFS pool, mirrored NVMe cache
Networking: 10GbE backbone, VLAN-segmented lab traffic
Hypervisor: Proxmox VE with container + VM workloads
Orchestration: Docker Compose, k3s for scheduled jobs
Backups: Snapshot schedule + off-site encrypted copy

Flagship Gaming Rig

A custom water-cooled tower built for high-refresh RPG rendering and heavy local workloads — spec'd part by part, tuned and benchmarked.

Dual-Purpose Workstation

The daily driver: analytics and marketing work by day, campaigns and creative writing by night — one desk, two very different workloads.

Talk shop?

Parts lists, thermals, benchmarks or campaign notes — I'm always happy to compare builds.

Get in touch Back to the hub