What does '99.9% availability' mean, and how much downtime is that?
99.9% availability, called three nines, allows about 8.76 hours of downtime a year, roughly 43 minutes a month. Each extra nine cuts the allowed downtime by ten times and costs much more to achieve.
Explain it simply
Availability is the fraction of time a service is working and able to answer requests. It is quoted as a percentage, and people say "nines".
Take a year of 365 days, which is 525,600 minutes. The allowed downtime is simply the remaining percentage of that. Think of an ATM: if it is out of service a few minutes in a month you barely notice, but a full day offline is a headline.
Higher nines are expensive because you must remove every single point of failure: multiple servers, multiple zones, automatic failover, careful deploys. Also, availability of a chain multiplies. Two services in series, each 99.9%, give about 99.8% together.
Example
Allowed downtime per year:
99% -> 1% of 525,600 = 5,256 minutes (about 3.65 days)
99.9% -> 0.1% = 525.6 minutes (about 8.76 hours)
99.99% -> 0.01% = 52.56 minutes
99.999% -> 0.001% = 5.26 minutesA payment gateway promising 99.99% has under an hour of yearly downtime to spend.
What to say in the interview
"Availability is uptime divided by total time. Three nines is about 8.8 hours a year, four nines about 53 minutes. I would explain that dependencies in series multiply, so the system is only as available as its weakest chain, and redundancy with health checks and failover is how we raise it."
Common mistakes
- Forgetting planned maintenance may or may not count; check the SLA wording.