Metrics for good, not evil
summary
Claims that SLOs/SLIs/SLAs create a neutral, data-driven bridge between product and engineering, with SRE owning telemetry and surfacing reliability vs feature tradeoffs. Concrete practices include defining 99.95% availability as an example, calculating a 22 minutes error budget per month, and using start with two to three SLOs to iterate. Error budgets get spent deliberately to balance risk and innovation, SLO breaches trigger priority shifts not individual blame, and metrics must reflect user pain not proxy vanity.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:50What is Rekka Point
- 4:26Product vs Engineering
- 7:19The Cobra Effect
- 10:34Humans are bad at metrics
- 11:18How to build metrics
- 15:05Speed and reliability compromise
- 16:03What are SLAs
- 18:52Error Budgets
- 20:10Speed vs Stability
- 20:27What does good IT look like
- 22:06Role of SRE
- 22:55Selling points
- 23:41Not for everyone
- 24:32Failure
- 25:30Blameless mentality
- 27:57Outro
related talks
38mImplementing SRE practices: SLI/SLO deep dive
29mMeasuring Reliability: What Got Us Here Won't Get Us There
30mTechniques for SLOs and Error Budgets at Scale
30mImproving Reliability w/ Error Budgets & SRE
09mHow to answer tricky questions the SRE way
35mAvoiding Goodhart's Law - Use SLO's as Tools Not Cudgels
47mCultivating Production Excellence
06ma pint sized introduction to SLO