Read the allowance over the whole reporting window
Allowed downtime is a total for the selected period. If a 30-day month has a 99.99% target, two separate outages of 3 minutes each use 6 minutes against a 4-minute-19.2-second allowance. The budget has been exceeded by 1 minute 40.8 seconds. The allowance does not restart after each incident.
The Google SRE availability table describes this relationship between availability targets and time unavailable. This calculator uses fixed 24-hour days: month means 30 days, year means 365 days. Choose Custom days for a 28-, 29- or 31-day month, a 366-day leap year, or another window. Do not compare a full-year outage total with a monthly allowance.
What the extra nines buy
These values come from the same time-based model. All durations below use the specified fixed periods, without intermediate rounding.
| Availability target | Downtime per day | Per 30-day month | Per 365-day year |
|---|---|---|---|
| 99% | 14 min 24 s | 7 h 12 min | 3 d 15 h 36 min |
| 99.9% | 1 min 26.4 s | 43 min 12 s | 8 h 45 min 36 s |
| 99.99% | 8.64 s | 4 min 19.2 s | 52 min 33.6 s |
| 99.995% | 4.32 s | 2 min 9.6 s | 26 min 16.8 s |
| 99.999% | 0.864 s | 25.92 s | 5 min 15.36 s |
Each move from 99% to 99.9%, or from 99.9% to 99.99%, divides the permitted downtime by ten. The move from 99.99% to 99.995% halves it. A 100% target gives a zero downtime budget; entering it does not predict that a service can deliver uninterrupted operation.
Calculate measured uptime from an outage total
Suppose a server has 10 minutes of downtime during a complete 30-day reporting period. The period contains 43,200 minutes, of which 43,190 were available:
Against a 99.9% target, the 43-minute-12-second allowance still has 33 minutes 12 seconds remaining. Against 99.99%, the same outage total exceeds the allowance by 5 minutes 40.8 seconds. A high-looking uptime percentage can therefore still miss a demanding target.
Enter the combined outage duration in Measured uptime mode and choose its unit. Switching from minutes to seconds changes how the number is interpreted: 10 minutes becomes 10 seconds unless you also change the number. Merge overlapping outages first. For example, incidents from 10:00–10:10 and 10:05–10:15 represent 15 minutes of service unavailability, not 20.
Uptime formula and downtime budget
Let T be the eligible reporting time, D the observed downtime and A the target percentage. Use the same unit for T and D.
Positive remaining budget is unused allowance. Zero means exactly exhausted. A negative value is the amount by which downtime exceeds the target budget. The calculation compares unrounded values: a tiny overrun still exceeds the budget even if the displayed uptime rounds to the target. Durations display to the nearest millisecond, with a special label for smaller nonzero values; percentages display up to six decimal places.
Match the metric to your SLA
Use this as a server or network availability calculator when availability is measured by time. It does not monitor a service or determine credits. An agreement may define maintenance exclusions, eligible hours, minimum outage lengths or other measurement rules. Apply those rules consistently to both the reporting window and downtime; this tool makes no automatic exclusions.
An internal availability target is an SLO. An SLA also defines consequences for missing its objectives, as explained in Google SRE's service level terminology. The numerical comparison here cannot decide those consequences.
Time-based uptime can also hide partial failures: one region or a subset of requests may fail while other traffic succeeds. Google SRE discusses request-based availability as a useful alternative for such services. Do not treat an elapsed downtime percentage as a failed-request percentage. The same time ratio can describe machine uptime, but it does not calculate production efficiency, MTBF or repair forecasts.
Read the formula definitions, supported input ranges, and limitations before using this estimate.
