Defining SLIs and SLOs for Platform Services
A practical guide to selecting indicators, setting targets, and reporting error budgets.
Defining SLIs and SLOs for Platform Services
SLI examples
| Service | SLI | Measurement |
|---|---|---|
| API | Availability | Successful requests / total |
| API | Latency | p99 < 300ms |
| CI pipeline | Success rate | Green builds / total |
| Auth | Login success | Valid logins / attempts |
SLO template
Over a 30-day window, 99.9% of API requests will succeed (non-5xx).Error budget = allowed unreliability = 0.1% ≈ 43 minutes downtime/month.
Policy
When error budget is expleted:
- Freeze non-critical releases
- Prioritize stability and security patches
- Resume features when budget recovers
Dashboards
- Burn rate alerts (fast budget consumption)
- SLO compliance by team / service tier
- Link incidents to budget impact in post-mortems
Security tie-in
Include security SLIs: mean time to patch critical CVEs, % prod workloads on signed images, MFA coverage for admin roles.