April 2023 — on making the network visible, organised, and not an embarrassment

Every home network starts the same way: a router from the ISP, a switch from a supermarket, a tangle of cables behind the television. It works. It’s not organised. It grows by accretion. At some point it becomes something you hide rather than show, and you stop understanding what’s actually connected to what.

The network shelf is the moment you stop hiding it and start being deliberate about it.

What lives on the shelf

MikroTik hEX (the white box with the blue LED): the router. MikroTik runs RouterOS, which is a serious routing OS that treats you like an adult. You configure it properly — firewall rules, static routes, VLAN tagging, DHCP reservations — and it does exactly what you configured. It doesn’t phone home, doesn’t have a cloud dashboard you need to register for, and doesn’t stop working when some cloud service has an outage.

The downside of MikroTik is that RouterOS requires you to know what you’re doing. There’s no wizard that sets up a sensible firewall by default. The default configuration on a factory-reset hEX is not secure. You have to understand what you’re configuring. This is a feature, not a bug — I want to understand what my network is doing.

TP-Link managed switches: the workhorse connectivity. Two TP-Link switches handle the device connections — one for the main shelf, one downstream. They’re managed (VLAN-aware, SNMP-capable) without being expensive or requiring a subscription to manage. They report into the MikroTik for network visibility.

The small circuit breakers and distribution: visible at the bottom of the shelf — a proper distribution panel for the equipment, not a power strip with everything plugged in at an angle. Each device gets a clearly labelled circuit. Power cycling one device doesn’t disturb the others.

The 10-inch open frame rack

Standard 19-inch rack equipment is too large for most home network cupboards and too heavy for most shelves that aren’t reinforced. The 10-inch (254mm) rack form factor exists for exactly this use case: small, light, designed for home network gear that’s built to that form factor.

The 10-inch open-frame rack is just two vertical rails with threaded holes at standard spacing. No side panels, no doors, no weight. It sits on a shelf or bolts to a wall, and equipment mounts to it with standard cage nuts and rack screws.

The equipment choices available in 10-inch form factor have expanded significantly: patch panels, keystone panels, 1U switches, 1U UPS units. Enough to build a complete, organised home network stack in a form factor that fits inside a 600mm-deep cupboard.

The open frame matters in a confined space: you can cable from any direction, and airflow isn’t restricted by side panels.

The shelf itself

The shelf is custom wood — the meter cupboard in a Dutch apartment has standard dimensions that don’t match anything commercial. A piece of solid wood, sized to fit, mounted on wall brackets. The network gear sits on the wood; the 10-inch rack bolts to the wall behind it.

Wood is fine for this. It doesn’t need to be a fabricated steel shelf. It needs to hold the weight (the gear is not heavy), allow cable routing behind and below, and look like it was intended, not improvised.

What the organisation actually changes

When the network is visible and labelled, you can reason about it. Which switch port goes to which device. Which VLAN carries which traffic. Which device is drawing power from which circuit. When something stops working, you can follow the physical path from device to switch to router and identify where the failure is.

When the network is a pile behind furniture, you can’t. You reboot things until something changes, and you have no idea whether the problem was the device, the cable, the switch port, or the router configuration.

Organised infrastructure is not cosmetic. It’s how you maintain and troubleshoot things that need to keep running.

The monitoring

The MikroTik exposes SNMP; the TP-Link switches expose SNMP. Prometheus with the SNMP exporter scrapes both. Grafana shows interface traffic per port, switch CPU and temperature, uptime, and the usual network health indicators.

This is the same monitoring stack that runs for the Kubernetes cluster and the IoT sensors. Everything goes into the same Prometheus instance. The home network gets the same observability treatment as anything else running in the house.

If a switch port goes to 100% utilisation, I know before anything breaks. If a device disappears off the network, the alert fires. If the router restarts, the uptime chart shows it.

The network isn’t a black box. It’s infrastructure. It gets instrumented like infrastructure.