# Building a Community Mesh-Network ISP

Our neighborhood ISP began with a practical constraint: several streets had poor broadband service, but no single household could justify the cost of a dedicated fiber build. We formed a cooperative, leased a 10 Gbps connection from a regional carrier, and installed the first point of presence in a community center with backup power and rooftop access. Members pay a modest installation fee and a monthly rate that covers transit, equipment replacement, insurance, and tower leases.

## Network Architecture

The access network uses rooftop routers connected through a mix of 60 GHz links and 5 GHz fallback paths. Larger relay sites form the mesh backbone, while individual homes usually connect as leaf nodes to avoid unnecessary routing complexity. We run Babel for dynamic route selection because it responds quickly to changing wireless conditions without requiring every volunteer operator to understand link-state tuning.

```conf
interface mesh0
interface mesh1
redistribute local deny
redistribute ip 10.42.0.0/16 allow
```

Each member router receives an address from a private infrastructure range and delegates a public IPv6 prefix to the home network. IPv4 traffic passes through carrier-grade NAT at the gateway, although members can request a public address for hosting services. Customer traffic is isolated with VLANs, and management interfaces are reachable only through WireGuard using cooperative-issued credentials.

## Operations and Reliability

Wireless links are monitored for latency, packet loss, signal level, and channel utilization. Prometheus collects metrics from relay nodes, while alerting rules distinguish between a failed router and a degraded path that Babel has already routed around. The most common incidents are power outages, trees growing into line of sight, and rooftop equipment shifting after severe wind—not failures of the routing protocol itself.

## Lessons from the First Year

The difficult part was less about radio technology than maintaining clear ownership. Every relay site now has a named steward, documented access instructions, photographs of cable routes, and a spare-parts kit. We also publish service expectations honestly: the mesh can survive individual link failures, but the shared upstream and core gateway remain critical dependencies until we can fund a second point of presence.