Measuring Reliability: What Got Us Here Won't Get Us There
summary
Argues that SLI/SLO metrics are useful but incomplete: a company with 200 SLOs sees Jan 3, Feb 4, Mar 12 violations and naive interpretation hides a 95% CI of ~3–13, so big decisions risk being wrong; error budgets are really business budgets (example $10,000/year) yet estimating outage impact is order-of-magnitude noisy. Recommends using diverse signals (logs, monitoring, social posts, customer reports), and a workflow to identify your question, build a model, and back-test it before acting.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:22What is reliability
- 3:33What is the DLDR
- 8:31First example
- 10:01Second example
- 21:53Understand your needs
- 23:07Operationalization
- 25:28SLO alerting
related talks
30mMetrics for good, not evil
38mImplementing SRE practices: SLI/SLO deep dive
30mTechniques for SLOs and Error Budgets at Scale
09mHow to answer tricky questions the SRE way
37mRelia...bility?
47mCultivating Production Excellence
32m9 Things You Should Do When Starting to Use SLOs
37mReal World SLOs and SLIs: A Deep Dive