Tools Updated 27 Oct 2026

Error budget and burn-rate calculator

Enter an SLO and window to see the error budget as time and as requests. Add the current error rate to see the burn rate, how long the budget lasts, and when each standard burn-rate alert would fire.

Alert thresholds for this SLO

Each row alerts when both windows exceed the burn rate. Detection time is when the long window crosses the threshold if the error rate jumps from zero and stays there with steady traffic; real alerts also wait for scrapes, rule evaluation and notification.
Action	Budget spent	Long / short window	Burn rate	Error rate threshold	Detection time
Page	2%	1h / 5m	14.4×	1.44%	—
Page	5%	6h / 30m	6×	0.6%	—
Ticket	10%	3d / 6h	1×	0.1%	—

How the calculator works

Error budget

error budget = 1 − SLO

A 99.9% SLO has an error budget of 0.1% of valid events in the window: 50,000 failed requests out of 50 million. With steady traffic that equals 43m 12s of complete outage over 30 days. Traffic is rarely steady, so an outage at peak spends more of a request budget than the same outage at night. Many small failures and one large outage both count.

Burn rate

burn rate = observed error rate ÷ (1 − SLO)

A burn rate of 1 spends exactly the whole budget over one window. At 99.9%, a 0.5% error rate is a burn rate of 5. Starting from a window with no errors, that exhausts a rolling 30-day budget in 6 days (window ÷ burn rate). Below a burn rate of 1 the budget never runs out: old errors leave a rolling window as fast as new ones arrive, and usage settles at the burn rate, for example 50% of the budget at 0.5×. Both assume steady traffic.

Alert thresholds

burn-rate threshold = budget share × SLO window ÷ alert window

The table uses the parameters recommended in Google’s SRE Workbook: page if 2% of the budget is spent in one hour or 5% in six hours, and open a ticket if 10% is spent in three days. Each long window has a short window of one twelfth of its length that must also be burning, so the alert resets quickly after recovery. For a 30-day window that gives the familiar burn rates 14.4, 6 and 1. Shorter windows give lower thresholds; a 7-day window pages at a burn rate of 3.36 over one hour. These parameters assume a window of about 30 days. For much shorter windows some thresholds drop below 1× and would alert while the service is still within its SLO, and an alert window as long as the SLO window makes no sense; the table marks both.

Read the multiwindow burn-rate alerting guide for the reasoning and Prometheus rules, and the error budget policy guide for what to do when the budget runs low.

Source: Google, The Site Reliability Workbook, ch. 5 “Alerting on SLOs”.