Examples
Examples
Try the starting example
- Contiguous equal-duration buckets in chronological order: label, good events, total events
- T1,1000,1000 T2,980,1000 T3,970,1000 T4,1000,1000
- Supplied SLO good-event target (%)
- 99
- Long window, whole buckets
- 2
- Short window, whole buckets
- 1
- Both-window burn threshold, strict greater-than
- 2
- Multiwindow alert state at each bucket end
- T1: insufficient history; T2: clear; T3: ALERT; T4: clear
- Long-window burn rate at each bucket end
- T1: Unknown; T2: 1; T3: 2.5; T4: 1.5
- Short-window burn rate at each bucket end
- T1: Unknown; T2: 2; T3: 3; T4: 0
- Bucket ends firing the modeled alert
- T3
- Number of modeled firing bucket ends
- 1
Try another scenario
- Contiguous equal-duration buckets in chronological order: label, good events, total events
- T1,1000,1000 T2,980,1000 T3,970,1000 T4,1000,1000
- Supplied SLO good-event target (%)
- 99
- Long window, whole buckets
- 2
- Short window, whole buckets
- 1
- Both-window burn threshold, strict greater-than
- 3
- Multiwindow alert state at each bucket end
- T1: insufficient history; T2: clear; T3: clear; T4: clear
- Long-window burn rate at each bucket end
- T1: Unknown; T2: 1; T3: 2.5; T4: 1.5
- Short-window burn rate at each bucket end
- T1: Unknown; T2: 2; T3: 3; T4: 0
- Bucket ends firing the modeled alert
- None
- Number of modeled firing bucket ends
- 0
Calculation method
Calculation method
Multiwindow alert state at each bucket end
Require full long-window history and positive traffic in both windows; fire only when both exact burn ratios strictly exceed threshold.
Long-window burn rate at each bucket end
Window bad-event rate/(1−SLO target), aggregating event counts rather than averaging per-bucket rates.
Short-window burn rate at each bucket end
Same burn ratio for the supplied short window.
Bucket ends firing the modeled alert
List endpoints where both-window condition is true, not continuous real-time alert periods.
Number of modeled firing bucket ends
Count firing evaluated endpoints; insufficient history and no traffic do not imply an alert.
Symbols and units
- a
- Contiguous equal-duration buckets in chronological order: label, good events, total events
- b
- Supplied SLO good-event target (%)
- c
- Long window, whole buckets
- d
- Short window, whole buckets
- e
- Both-window burn threshold, strict greater-than
Use the units shown. A list keeps the order in which you enter its values. The symbol ^ means “raised to the power of”.
Assumptions & limits
Assumptions & limits
- An endpoint trace under supplied contiguous equal-duration count buckets and strict two-window thresholds. It is not a Prometheus evaluator, live alert, interpolated continuous timeline or proof that buckets were actually complete.
Are these inputs private?
These calculations run in your browser. A shared link contains the values you choose to share. Only questions submitted to the separate AI finder are sent to OpenAI.
Further reading
Google SRE Workbook: multiwindow burn-rate alerts ↗
Background reading; this does not imply endorsement or independent review of SumFizz.