Latency and Availability Error Budgets Done Right at Scale
summary
Zendesk codifies SLI → SLO → error budget as simple math: metric identifier operator metric value and success objective over a time window, then error budget = 1 − SLO. Implementation runs in Datadog with 7/30/90 day windows, 99.5% target, and alerts when ~25% of a budget burns. At scale (≈1,000 engineers across 150 teams) the playbook mandates consistent templates and, for mid‑tier services, two error budgets per service to separate internal vs dependency contributions (examples show 1.2% and 1.3% total burn).
AI-written summary · about
chapters (7)
- 0:00Intro
- 1:14Air Budget Refresher
- 4:15Service Level Indicators
- 8:43Error Budget Formula
- 10:14Simple Examples
- 10:53Example
- 13:07Managing air budgets across multiple services
related talks
30mReliable Observability at Scale: Error Budgets for 1,000+
30mTechniques for SLOs and Error Budgets at Scale
30mMetrics for good, not evil
17mRolling out Service Levels across a large org
36mHow to Include Latency in SLO-based Alerting
09mHow to answer tricky questions the SRE way
38mImplementing SRE practices: SLI/SLO deep dive
37mRelia...bility?