SLI
What to measure. Availability: successful requests / total. Latency: p99 < 200ms. Correctness: valid responses.
Advertisement
SLO
Target: 99.9% availability over 30 days. Concrete + measurable. Basis for engineering decisions.
Advertisement
Error budget
100% - SLO = allowed failures. 99.9% = 43.2 min downtime/month. Spend on features or reliability.