On-Premises (aka, in my apartment)
Below is the list of all of the compute nodes I am running in my apartment.
Mamamoo cluster
Back in 2023, I bought 4x RockPI 4A+ (read about this endeavor here). This was the first expansion to my homelab, bringing redundancy and high availability to my most commonly used services. They are all plugged in to a Netgear GS324TP, which is supplied power via a CyberPower CP1500PFCLCD.
Moonbyul
and Wheein
are responsible for low-requirement, high-throughput tasks (e.g. rarely-used webapps, text processing, DNS), while Hwasa
and Solar
provide low latency and redundancy in data access, for applications that require high network throughput.
members
All nodes are identical except the storage device attached. They are RockPi 4A+ with PoE HAT and M.2 extension board.
- Moonbyul: silicon power P34A60 (256GB)
- Wheein: silicon power P34A60 (256GB)
- Hwasa: silicon power P34A60 (1TB)
- Solar: silicon power P34A60 (1TB)
I.O.I cluster
For a multitude of reasons, I've come into the possession of some decently-spec'd laptops. Rather than have them sit in a closet, unused, I use them for background tasks that require decent compute and are tolerant of slow network speeds. Most of their tasks revolve around data visualization and preprocessing input for AI.
members
- Sejeong: razer blade 15
- Chungha: dell xps 9510
- Somi: dell precision 5570
soloists
I also have a couple PCs built from consumer hardware. These nodes are reserved for tasks that require high CPU, GPU, or network throughput.
members
- IU, a small form factor PC (intel i9-13900k, 128gb ram, rtx 3090 ti), used mainly for AI workloads
- Jessi a PC I salvaged for free. Used as a NAS device.