Designing for the link you cannot drive to

Satellite teaches habits that terrestrial networks let you skip. Most of them are worth keeping once you come back down.

On a terrestrial network, a bad decision is cheap. Console into the switch, back the change out, move on. On a satellite remote, the same mistake can mean a site is down until somebody physically travels there. That asymmetry changes how you work, and the changes are all improvements.

Observability before convenience

When you cannot walk over and look, the only thing standing between you and a site visit is the data you arranged to collect beforehand. Monitoring is installed before the site is handed over, not after the first outage. Every remote reports enough to diagnose it without touching it — and the time to discover a gap in that telemetry is during commissioning, not during a fault.

Terrestrial networks let you postpone this, so people do.

Configuration is validated before it ships

A remote terminal is configured, tested and verified before it leaves. Not because the process is elegant but because the alternative is unacceptable. The same discipline applied to a branch switch — staged, configured and tested on the bench before it is couriered to a site with no technical staff — eliminates an entire category of deployment day problems.

Degradation is a design state, not an incident

Rain fade is not a fault. It is a condition the link will enter regularly, and the design either accounts for it or it does not. Margin is calculated, behaviour under degradation is known, and the system is expected to keep working at reduced capability.

Terrestrial designs rarely define what degraded looks like. They define working and failed, and then behave unpredictably in between. Asking what happens when this link is at half capacity rather than down is one of the more useful questions in a design review.

Scarcity forces prioritisation

Space segment is expensive enough that you cannot avoid deciding what matters. Traffic gets classified and shaped because the alternative is a backup job starving a voice call. Terrestrial bandwidth is cheap enough that these decisions get deferred — until the day the link is saturated and nobody has decided what should win.

The backup path has to be real

Hybrid designs with satellite as the backup fail in one specific way: the backup has never carried production traffic, so nobody knows whether the applications work across it. Latency alone breaks assumptions that terrestrial links let applications make.

Which is the general rule, and worth stating plainly: a redundant path that has never been used is not redundancy. It is an assumption. Test it by switching the primary off, on purpose, in a window you chose.