In my homelab journey, I’ve ended up with two very different kinds of clusters:

A central cluster in Ontario (running on beefier Proxmox/Ceph hardware).

A portable edge cluster in Newfoundland (powered by tiny NanoPi Neo boards and a travel router).

At first glance, they couldn’t be more different. One has cores to spare, terabytes of SSD storage, and sits in a proper rack. The other runs on dusty SBCs with half a gig of RAM each, powered off USB battery packs.

But that’s the beauty of homelabbing: each cluster brings something different to the table.

The Central Cluster

The central homelab is where the “heavy lifting” happens. It’s always online, has redundancy built in, and can host services that require consistency and scale.

Best fits for the central cluster:

Storage & Backups – Ceph-backed storage, Proxmox backups, and media servers.

Virtualization & Orchestration – Proxmox VMs, Kubernetes workloads, multi-node databases.

Long-running Services – Plex, Nextcloud, Valkey clusters, monitoring stacks (Grafana, Prometheus).

CI/CD & Registries – central Git repos, Docker registries, and pipelines.

High-resource Apps – anything that needs multiple cores, lots of RAM, or fast disk I/O.

The central cluster is essentially the datacenter backbone of the homelab.

The Edge Cluster

The NanoPi Neo mini-cluster is the opposite: low-power, lightweight, portable. But that makes it perfect for a different set of scenarios.

Best fits for the edge cluster:

Testing & Dev Environments – spin up experimental apps without risking central stability.

Local Services – run a blog, a small DB, or monitoring tools while offline.

Battery-Powered Experiments – resilience tests, low-power compute studies, and field deployments.

Data Collection – logs, metrics, or IoT-style workloads that sync back to the main homelab later.

Distributed Experiments – test software designed to run across multiple nodes, even if they’re tiny.

The edge cluster is essentially the lab-on-the-go — lightweight but flexible enough to get real work done in remote places.

Why Both Matter

The real power comes when the two clusters work together:

The edge cluster handles local collection, testing, and short-lived workloads.

The central cluster provides storage, redundancy, and long-term hosting.

Tailscale stitches them into a single Tailnet, so data and workloads can flow back and forth securely.

This creates a homelab that’s both resilient and flexible: centralized for scale, distributed for mobility.

Closing Thoughts

Not every workload belongs in the same place. By pairing a central cluster with an edge cluster, I get the best of both worlds: power at home, agility on the road.

It’s not about which is better — it’s about how they complement each other. In the end, both clusters contribute something unique to the conversation in the homelab.