Two ways to think about downtime
Downtime calculations go in two directions. Forward: given an SLA target, how much downtime is permitted? Reverse: given actual downtime that occurred, what uptime percentage does that represent? Both calculators above cover each direction.
Downtime budget by SLA tier
| SLA target | Annual downtime budget |
|---|---|
| 99% | 87h 36m |
| 99.5% | 43h 48m |
| 99.9% | 8h 45m |
| 99.95% | 4h 22m |
| 99.99% | 52m 35s |
| 99.999% | 5m 15s |
See the full breakdown including per-week, per-day, and per-hour figures in the Uptime Percentage Table.
Tracking cumulative downtime
SLA downtime budgets are cumulative across the measurement window — multiple short outages count the same as one long outage of equal total duration. A service with three 3-hour outages in a month has used 9 hours of downtime budget, identical to a single 9-hour outage.
This is why tracking downtime incrementally throughout the window matters more than checking only at the end. Use the Error Budget Calculator to track burn rate in real time as incidents occur.
What counts as downtime?
Most SLAs define downtime narrowly — typically complete unavailability or error rates above a defined threshold (e.g. more than 5% of requests failing). Partial degradation, elevated latency without errors, and individual feature outages often fall outside strict SLA downtime definitions even though users may experience them as "the service is down."
Always check your specific SLA's definition of downtime before assuming an incident counts against the budget.