The “nines” of availability
Availability is the share of time a service works. Service level agreements (SLAs) usually express it as a number of nines, and each extra nine divides the allowed downtime by ten.
| Availability | Per year | Per month | Per week | Per day |
|---|---|---|---|---|
| 99% | 3 d 15 h 39 min | 7 h 18 min | 1 h 41 min | 14 min 24 s |
| 99.9% | 8 h 45 min 58 s | 43 min 50 s | 10 min 5 s | 1 min 26 s |
| 99.95% | 4 h 23 min | 21 min 55 s | 5 min 2 s | 43 s |
| 99.99% | 52 min 36 s | 4 min 23 s | 1 min | 8.6 s |
| 99.999% | 5 min 16 s | 26 s | 6 s | 0.9 s |
The calculator uses a 365.25-day year and an average month of 30.44 days.
Systems with several parts
Real services depend on several components. The calculator combines them in two ways:
- In series, every part must work, so availabilities multiply: three components at 99.9% give 0.999³ = 99.7%, about 26 hours of downtime a year.
- Redundant copies in parallel fail only when all fail at once: two independent 99.9% copies give 1 − 0.001² = 99.9999%.
That is why architects add redundancy to the weakest links. The formula assumes failures are independent; shared power, networks or software bugs make real redundancy less effective.
SLA tips
- Check how the provider measures downtime: per minute, per request, excluding planned maintenance or not.
- SLA credits are usually a small share of the monthly bill, not compensation for lost business.
- Your service cannot be more available than the product of the dependencies it needs in series.
- Track your own error budget: the allowed downtime is a budget you can spend on deployments and experiments.