How to use this calculator
- Enter or choose sLO target.
- Enter or choose expected events in the window.
- Enter or choose failed events so far.
- Enter or choose sLO window.
- Enter or choose days elapsed.
- The error budget (failed events allowed) updates as you type. Use Copy result or Share to keep it.
Method
Error budget = (1 − SLO) × Total events; Consumed = Failed ÷ Budget; Burn rate = Consumed ÷ (Elapsed ÷ Window)
- SLO
- target success ratio
- Burn rate
- 1 means the budget runs out exactly at the end of the window
Worked example
- Budget: 0.001 × 10,000,000 = 10,000 failed events.
- Used: 4,000 ÷ 10,000 = 40% after 10 of 30 days (33.3% of the window).
- Burn rate: 0.40 ÷ 0.333 = 1.2, so the budget runs out in about 15 more days.
A 99.9% SLO on 10 million events allows 10,000 failures; using 4,000 in the first third of the window is a burn rate of 1.2.
What the result means
A burn rate above 1 means the budget will be gone before the window ends if nothing changes. Many teams alert on short, high burn rates (fast outages) and long, moderate ones (slow leaks), and slow risky releases while the budget is low.
Assumptions and limits
- Traffic is spread evenly over the window, so elapsed time is a fair measure of expected events so far.
- Time-based SLOs (minutes of downtime) use the same idea with minutes instead of events; see the Uptime SLA Calculator.
Questions
Where does the idea come from?
Error budgets and burn-rate alerting are described in Google’s Site Reliability Engineering books.
Is burn rate the same as error rate ÷ (1 − SLO)?
Yes, when traffic is even: the observed error rate divided by the allowed error rate gives the same figure.