The nines & your error budget
What each level of uptime actually costs you in downtime, and the error-budget idea that turns reliability from an argument into a decision.
Google SRE
Reliability isn’t a vibe, it’s a budget. Google’s SRE discipline turns “how reliable should we be?” into a number you can spend. Below is the most screenshotted table in the field: what each level of uptime actually costs you in allowed downtime.
| Uptime | Per day | Per month | Per year | |
|---|---|---|---|---|
| 90% | one nine | 2.4 hours | 3 days | 36.5 days |
| 99% | two nines | 14.4 minutes | 7.2 hours | 3.65 days |
| 99.9% | three nines | 1.4 minutes | 43.8 minutes | 8.77 hours |
| 99.95% | 43 seconds | 21.9 minutes | 4.38 hours | |
| 99.99% | four nines | 8.6 seconds | 4.4 minutes | 52.6 minutes |
| 99.999% | five nines | 0.86 seconds | 26.3 seconds | 5.26 minutes |
Read it out loud: 99.9% uptime means you’re allowed about 44 minutes of downtime a month.That’s your error budget. Each extra nine costs roughly 10× the engineering effort, so the honest question is never “how reliable can we be?” but “how reliable does this actually need to be?”
The four words people mix up
A service level indicator. One number that measures a thing users care about, usually good events ÷ total events.
A service level objective. The target you set for an SLI. This is the line the team commits to, internally.
A service level agreement. The promise you make to customers, with consequences. Always looser than your SLO.
Simply 1 − SLO. If your SLO is 99.9%, you are allowed 0.1% of failure. Spend it on shipping; when it runs out, you freeze and stabilise.
An error budget turns reliability from an argument into a decision. While the budget has room, ship freely. When it’s gone, the same rule binds everyone: stop shipping features and spend the effort on stability until you’re back in budget. No hero, no blame, just the maths. (SRE also targets keeping toil, manual, repetitive, automatable work, under 50% of an engineer’s time.)
Concepts from Google’s Site Reliability Engineering practice (the free O’Reilly SRE books). Explained in original words; downtime figures are arithmetic from each uptime target.
Want the version tuned to your actual situation?
Book a call