Reliability in our markets is not a network problem. It's a power problem, a logistics problem, and a procurement problem. The network is the easy part.
Design for the worst day
We size battery backup for 18 hours, not 4. We pre-position spares within 200km of every PoP. We assume the WAN link will fail and that we'll route over LTE for a week. None of this is exotic engineering — it's the discipline of building for the third-worst day, not the best one.
The metric we actually optimize
Not uptime. Time-to-recovery. Outages will happen. Customers forgive 30 minutes; they leave at 4 hours. Every internal process we have is tuned around shrinking the gap between detection and restoration.
