Observability: it's not just an ops thing
summary
Argues that traditional monitoring (alerts + dashboards) only answers known questions and that true observability requires scuba-like high-cardinality querying and fast, ad-hoc exploration. Describes building a custom storage engine for sub-second queries at 95th percentile, shipping one rich event-per-request with many attributes, and using sampling so rare cases are retained. Claims this cuts debug time from hours/days to seconds/minutes, drives decisions like making JSON unfurling opt-in after traffic analysis, and replaces noisy alerting with a few end-to-end guards and deploy triggers.
AI-written summary · about
chapters (4)
- 0:00Start
- 3:18MONITORING IS DEAD
- 13:05Software Development Process
- 22:39Observability-Driven Development
related talks
44mObservability for Emerging Infra: What Got You Here Won’t Get You There
13mObservability Where are We Headed?
42mObservability Is Not Analytics!
43mObservability for Emerging Infra: What Got You Here Won't Get You There
29mObservability; a gentle introduction
31mObservability for Developers: How to Get There
22mObservability is Too Expensive!
32mEssential Software Observability for Every Role: Look No Further