# Building a Community Mesh-Network ISP

When unreliable broadband became a recurring problem in our neighborhood, we built a small community-operated ISP instead of waiting for another commercial provider. The network now serves 180 households across roughly six square kilometers using rooftop radios, volunteer-hosted relay sites, and two leased fiber uplinks. Members pay a monthly fee that covers transit, replacement hardware, insurance, and a modest reserve for expansion.

## Network Architecture

Most homes connect through 5 GHz point-to-multipoint radios aimed at one of twelve rooftop distribution nodes. Those nodes form a redundant 60 GHz mesh, with selected sites connected to the fiber gateways. We use OSPF internally because the topology changes often enough that static routes became difficult to maintain. Each household receives an isolated VLAN, while management traffic travels over a separate network accessible only through WireGuard.

A typical distribution-node interface is configured like this:

```ini
auto vlan204
iface vlan204 inet static
    address 10.24.4.1/24
    vlan-raw-device mesh0
    mtu 1500
```

## Capacity and Reliability

Wireless links are planned around measured airtime rather than advertised radio throughput. We target no more than 60 percent peak channel utilization and move heavy users between sectors when latency begins rising under load. Gateways run CAKE queue management to control bufferbloat, and Prometheus collects signal strength, packet loss, interface errors, and per-sector utilization. If either fiber circuit fails, BGP shifts outbound traffic to the surviving provider within about a minute.

## Operations and Security

Running the network is closer to operating a utility than maintaining a hobby project. Firmware upgrades are staged on a test node, configuration changes are reviewed by two volunteers, and nightly backups are stored off-site. Subscriber devices cannot communicate directly with one another, and rooftop management interfaces accept connections only from the operations VPN. We also publish an incident log so members can see whether an outage came from weather, upstream transit, power loss, or our own mistakes.

## Lessons From the First Year

The hardest problems were access agreements, safe rooftop work, and support—not routing protocols. Standard mounting kits, documented installation checklists, and scheduled volunteer rotations reduced failures more than any software change. A community mesh ISP can provide capable local broadband, but it succeeds only when governance, maintenance funding, and operational discipline are treated as core infrastructure.