Developer & Tech

SRE Burn Rate Calculator

What this does

Compute how fast you are consuming error budget relative to plan, benchmarked against Google SRE fast-burn conventions.

Enter your details

Runs in your browser

Calculator inputs

How to use it

Using the sre burn rate calculator

  1. 01

    Enter SLO and window

    Same values as your error-budget dashboard so rates stay comparable.

  2. 02

    Enter observed errors

    Downtime minutes (or failed-request equivalents) accumulated since the window started.

  3. 03

    Compare against thresholds

    Above 14.4x page immediately; above 6x investigate same-day; between 1x and 6x watch the trend.

Good to know

Where 14.4x comes from

Burning 2% of a 30-day budget within one hour projects a 14.4x rate; small enough to catch real outages fast, large enough to ignore noise. The 6x threshold catches slower bleeds within a working day. Google published these constants in the SRE Workbook; teams tune them to their own risk tolerance.

Multiwindow pairing

Production alerting pairs a short window (fast detection) with a long one (false-positive suppression); e.g. fire only when both 1-hour and 5-minute burn exceed threshold. This calculator gives you the single-window rate those policies build upon.

How it's calculated

The math behind this calculator

burn = (observed ÷ window) ÷ (allowed ÷ window) = observed ÷ allowed conventions: 14.4x ≈ 2% of a 30-day budget in 1 h · 6x ≈ 5% in 6 h

Burn rate divides your observed error rate by the rate the SLO permits. A value of 1x means spending budget exactly on schedule; higher values project breach before the window closes. The 14.4x and 6x thresholds come from Google’s multiwindow alerting conventions and are labeled here as conventions rather than laws.

Assumptions & limitations

  • Errors and budget share the same units (minutes here).
  • Thresholds follow Google SRE Workbook multiwindow alerting conventions.
  • Steady-state burn assumed over the elapsed period.

Worked example

With a 99.9% SLO over 30 days the allowed budget is 43.2 minutes; observing 120 minutes burns at 2.78×; elevated but below the 6x paging threshold.

FAQ

Frequently asked questions

What burn rate is “normal”?
Exactly 1.0x consumes the budget perfectly across the window. Sustained rates under 1 leave headroom; anything persistently above 1 ends in breach.
Can I use requests instead of minutes?
Yes; divide failed requests by the request-denominated budget. The ratio is unit-independent.
Why not alert on raw error count?
Counts ignore context: 100 errors matter more at 99.99% than at 99%. Burn rate normalizes against what you promised.

Keep exploring

Related calculators