Cloud Service Measurement
Domain 5 tests whether you can measure service performance, risk, and value without misleading the business.
Core rules
- A metric is a measurement; a KPI is a metric tied to a target; a KRI is a risk indicator; an SLI is the measured behavior; an SLO is the internal target; an SLA is the contractual commitment.
- Use leading and lagging indicators. Availability, latency, and error rate are common SLIs; cost per transaction and adoption measure value.
- Report by audience: operations need detail, service owners need trends and breaches, executives need outcomes and risk.
- Measure end-to-end service, not only provider uptime. Include customer-side configuration, dependencies, and support performance.
Worked example
A 30-day month has 43,200 minutes. An SLO of 99.9% allows 43.2 minutes of downtime. The service records 50 minutes of unavailability, so it misses by 6.8 minutes. The team should not hide this with monthly averages; it should review the incidents, error budget, and customer impact. Reasoning: convert percentages into minutes and compare against the actual window.
Traps
Vanity metrics; averages that hide tail latency; measuring infrastructure only; no baseline or target; reporting SLA penalties instead of improvement actions; and ignoring unit cost and customer experience.