All books

Professional

Apps About Coach Log in Start reading

Quantitative Finance · Glossary

What is Burn-rate alert?

Definition 28.6 Research, Data and Risk Platforms · Chapter 28 — Observability and Incident Response

A burn-rate alert pages when the error budget is being spent faster than a stated multiple of the rate that would exactly exhaust it over the SLO’s period, measured over a long window and confirmed over a short one: for a 99.9% objective, Google’s SRE workbook recommends paging at 14.4 times over an hour (confirmed over five minutes) or 6 times over six hours (over thirty minutes).

rulefeed stallintermittent stallsslow consumerfalse pages a month
process heartbeatnevernevernever0
data age above 5 s5 s5 s100 s18
data age above 30 s30 snever600 s1
data age above 60 s60 snever1 200 s0
burn rate, 1 h and 6 h rules56 s191 s151 s1
Table 28.1. Time to detect each replayed incident, and pages over thirty quiet days, for five alerting rules. The heartbeat never fires; the static thresholds buy speed with false pages or miss the intermittent stalls; the burn-rate rules detect all three within about three minutes and page once a month on quiet days, for the day’s 56-second pause.
Read in context →