What 99.9% uptime actually means when the client is a hospital

Every hosting provider advertises 99.9%. Very few can tell you what they do in the ninth minute of an outage, which is the only part that matters.

99.9% monthly allows roughly 43 minutes of downtime. In a hospital, 43 minutes of unavailable admissions and billing is a genuine clinical and financial event — so the number alone is not a plan.

Automated monitoring is worthless if the alert lands in an inbox nobody reads overnight. Our checks — uptime, CPU, disk, database connection and backup success — escalate to an on-call engineer's phone.

The target is that we know before the client does. In practice that is what separates a fifteen-minute incident from a two-hour one.

An untested backup is a belief, not a safeguard. We run quarterly restore drills into a scratch environment and time them, because the first time you restore should never be during an emergency.

Backups are held off-site with 30-day retention, so a corruption discovered a fortnight later is still recoverable.

When we moved Green Life Care onto a new platform, we did it department by department over six weeks, running old and new in parallel with a rollback path at every step.

It was slower and more expensive than a single cutover weekend. It also meant that at no point could a failure stop the hospital from admitting a patient. For this class of client, that trade is not close.