The basic formula
Uptime percentage is the share of a measurement window during which a service was available. The formula is (total time − downtime) ÷ total time × 100. If a service was down for 45 minutes in a 30-day month — 43,200 minutes — its uptime is (43,200 − 45) ÷ 43,200 × 100, or about 99.90%.
The number is only as meaningful as the definition of "down" behind it. Whether a slow-but-responding service or a single failed region counts as downtime is a decision you make; state it, because it changes the result.
What the nines actually allow
People describe availability targets in "nines", and each nine is an order of magnitude stricter. The intuition is easier with allowed downtime than with percentages. Assuming a 30-day month, 99% uptime allows about 7.2 hours of downtime; 99.9% ("three nines") allows about 43 minutes; 99.99% ("four nines") allows about 4.3 minutes; and 99.999% ("five nines") allows about 26 seconds.
Across a full year the same targets allow roughly 3.65 days at 99%, 8.76 hours at 99.9%, 52.6 minutes at 99.99%, and 5.26 minutes at 99.999%. The jump from three nines to four nines is a tenfold reduction in permitted downtime, which is why each additional nine costs far more than the last.
Why the measurement window matters
A percentage is meaningless without a window. Ninety minutes of downtime is 99.79% over a 30-day month but only 99.98% over a year, so a provider quoting an annual figure can absorb an outage that would blow a monthly budget. This is why service commitments almost always specify the window explicitly, and why a monthly window is stricter and more honest for most services.
Shorter windows also reset more often, which cuts both ways: one bad afternoon dominates a monthly number, but you are never carrying a year-old outage against today's reliability.
Availability, MTBF, and MTTR
Two operational levers move uptime: how often things break and how fast you recover. Mean time between failures (MTBF) captures frequency, and mean time to recovery (MTTR) captures speed. Availability can be expressed as MTBF ÷ (MTBF + MTTR), which shows that cutting recovery time improves uptime just as much as preventing failures. For most teams, detecting and resolving incidents faster is the cheaper lever.
This is where fast detection pays off directly: minutes shaved off MTTR go straight into the availability number.
Where the numbers come from
To report uptime honestly you need an independent, outside-in record of when the service was up. External monitoring provides that timeline, and a status page can turn it into published uptime history for customers. Remember that your measured uptime depends on how often you check: a five-minute interval can miss a three-minute outage entirely, so tighter checks produce a more truthful number.