Error budget calculator

The short answer#

• The error budget is 1 − target of the window: 99.9 % over 30 days allows 43.2 minutes of bad time.

• Burn rate is how fast you spend it: a rate of 1 uses the whole budget exactly at the end of the window.

• Belvador pages at a burn rate of 14.4 over 1 hour (checked against 5 minutes) and 6 over 6 hours (checked against 30 minutes), and opens a ticket at 1 over 3 days (checked against 6 hours).

• Two windows per alert stop a short spike from paging and let an alert clear soon after the problem ends.

How the numbers are calculated#

• Allowed bad time = (1 − target) × window length.

• Failures per million = (1 − target) × 1,000,000, for request-based SLOs.

• Burn rate = observed error ratio ÷ (1 − target).

• Budget gone in = window length ÷ burn rate, if the rate stays the same.

The alert thresholds follow the multi-window, multi-burn-rate method in Google's SRE workbook, which is defined for a 30-day budget. Belvador applies the same factors to 7- and 28-day windows, so on a 7-day window a 14.4 burn exhausts the budget in about 12 hours instead of two days.

Choosing a target#

Start from what your users notice and what your dependencies allow. A service behind a provider with a 99.9 % SLA cannot promise 99.99 %. Targets above 99.99 % leave a budget of seconds per day, which most teams cannot act on; Belvador accepts targets from 90 % to 99.99 %.

Using it in Belvador#

An SLO can count uptime-check rounds, monitor runs, ping-URL runs or spans. Belvador evaluates it every minute and alerts through your destinations (SLOs, burn-rate alerts).