SLO burn-rate calculator
Compare recent and longer-window error rates against an SLO and your chosen burn-rate threshold.
Free to use. Save a file draft or explicitly enable storage on this device.
Current result
Enter the objective and both event counts.Incomplete or uncertain. Review the fields and warnings.
Review resultsStart with your data or try an illustrative example.
File drafts are available without browser storage. Current draft matches the last save or initial state.
Inputs & calculation
Enter the objective and both event counts.
Use a common end time so the short window describes the most recent part of the long window. Compare their rates; adding the overlapping windows would count some events twice. With little traffic, a few failures can dominate the result, so read the counts alongside the rates.
Results
Review and export
This is incomplete or has review issues. You can save a draft; exports retain unknowns and warnings.
SLO burn-rate calculator report
SLO burn-rate calculator User-entered draft; verify against source evidence. Tool version 2 · https://aiopssre.com/slo-burn-rate-calculator/ Name: Untitled Scope: SLI: Not specified SLO: 99.9% over 30 days. Enter the objective and both event counts. This is a threshold illustration, not a configured alert. Windows must end together and use the same eligible population and SLI. Normalized spend assumes steady traffic; actual remaining rolling budget is not measured here.
| Window | Eligible events | Bad events | Duration | Burn rate |
|---|---|---|---|---|
| Long | 1 hours | Invalid | ||
| Short | 5 minutes | Invalid |
How the burn-rate calculation works
Burn rate divides the observed bad-event fraction by the fraction allowed by the SLO. A 99.9% target permits 0.1% bad events, so an observed 1% is 10× burn. One hour at that rate is about 1.39% of a normalized 30-day budget. The estimate assumes steady traffic; actual consumption can differ when request volume changes.
Why there are two observation windows
A long window can remain elevated after errors subside. The shorter window shows whether the recent error rate also meets your threshold. In the illustrative dataset, both windows show 15× burn and exceed the 14.4× checkpoint. Enter counts for the same service indicator and a common end time so the short window sits inside the long one. Their overlap is intentional: compare the rates rather than adding their spend. The report tells you whether both rates meet the checkpoint; it does not configure an alert or measure the remaining rolling budget.
Read the accompanying guide · Source and further reading
Calculation and worksheet review: September 29, 2026.