SLIs, SLOs, and Error Budgets

A healthy service dashboard does not always mean customers are having a good experience. An orders API might return successful responses most of the day while customers report intermittent failures and slow screens. Healthy processes and low average latency leave a more useful question unanswered: how often can a customer actually complete an order lookup within an acceptable time?

Service level indicators turn that experience into measurements. Service level objectives set targets for those measurements, and error budgets make the remaining tolerance for failure explicit.

This chapter explains how to define them for an API and use them to guide operational decisions.