When an organisation quotes its uptime, it is really quoting the quality of its earlier design decisions. A single point of failure in the core switch, no documented recovery runbook, or monitoring that only fires after users complain — each of these quietly caps how reliable the system can ever be.
Reliability by default means assuming components will fail and designing around it: redundant paths, tested failover, backups that are actually restored in drills, and alerting that gives you minutes instead of a helpdesk ticket.
That is why I treat uptime as an architecture concern. Get the design right and the metric takes care of itself.
