Nines Table
The mapping between availability percentages and permitted downtime, and the cost curve that comes with it.
Per year and per month:
99% — 3.65 days / 7.3 hours. 99.9% — 8.76 hours / 43.8 minutes. 99.99% — 52.6 minutes / 4.4 minutes. 99.999% — 5.26 minutes / 26 seconds.
Two things follow immediately.
Each nine costs roughly an order of magnitude more than the last, and the mechanisms change qualitatively. 99.9% is achievable with redundancy and competent operations. 99.99% requires multi-zone deployment, automated failover and no manual step in recovery. 99.999% requires multi-region active-active and eliminates most human intervention — which usually means the change process, not the architecture, becomes the constraint.
Below 99.99%, human response cannot meet the target. Four minutes a month does not allow for paging, waking, diagnosing and acting. That is the boundary where automated recovery stops being an optimisation.
The productive question is therefore not "how many nines do we want" — the answer is always more — but what the next nine costs and what it is worth, quantified against the actual cost of downtime.