Skip to content

What is an SLO?

An SLO (Service-Level Objective) is a concrete reliability promise about a service — a measurable target and the window you measure it over. Instead of “the API should be fast,” you write “99% of API requests complete under 300 ms over the last 30 days” and let the data confirm it.

The rest of this page defines the vocabulary the Reliability area uses so the other pages read cleanly.

TermPlain-English meaning
SLO (Service-level Objective)A reliability promise written as “this service should achieve X% over a Y-day window.”
SLI (Service-level Indicator)The specific signal being measured — e.g. spans under 200 ms, seconds without an open incident, or a good/total ratio of custom metrics.
Target %The promise — e.g. 99.9%.
WindowThe look-back period for the calculation — a rolling 1, 7, or 30 days, or the current calendar month.
Compliance %Where the service is right now versus the target.
Error budgetHow much “bad” you can afford under the target before breaching. For the time-based SLIs (incident availability, time slices, synthetic uptime) it’s a duration of downtime; for the count-based SLIs (the APM kinds and custom metric) it’s a count of bad events.
Burn rateHow fast you’re consuming the error budget right now relative to the rate that would keep you exactly on target. means you’re on pace; 14.4× means you’re consuming a 30-day budget in a few hours.
Burn-rate alertA notification rule that fires when burn rate is too high over both a short and a long window simultaneously. The long window suppresses brief blips; the short window keeps the alert responsive.

An SLO is a target plus a window. Each evaluation produces a compliance % (the share of good events). The gap between compliance and target gets translated into an error budget — the count or duration of bad events the service can absorb before falling below target. Burn rate is the rate at which the budget is being consumed. A burn-rate alert pages on that rate before the budget is fully consumed.

Workspace admin is required to create, edit, or delete SLOs and burn-rate alerts. Everyone else can view all reliability surfaces; the create, edit, and delete actions are hidden for non-admins.