The Ticking Time Bomb of Observability Expectations
summary
Argues that “monitor everything” plus vendor-driven marketing creates runaway cost and noise: observability often consumes 20% of cloud bill, floods teams with billions of low-value events, and vendors respond by shrinking retention or raising prices. Recommends treating dashboards as a product, designing status vs diagnostic views, instrumenting one or two customer-centric metrics, routing telemetry through an OTel collector to strip high-cardinality tags, and notes real cases like a 10x bill spike from tag cardinality.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 1:36Problems
- 2:26Expectations
- 3:22Data
- 4:25Architecture Trends
- 5:25Flow of Money
- 6:37Cost Effective
- 7:22Cognitive Load
- 10:13Dead Ends
- 12:53Cost Pressure
- 13:41The Solution
- 13:51Constructing Meaning
- 15:49Construction Meaning
- 17:32How to Attack This
- 19:08General Tips
- 23:38Status Information
- 25:16Instrumenting in Plain English
- 26:55Cost
- 29:38Otel Collector
- 30:13Assumptions
- 32:23How do you know
- 32:50Questions to ask
- 34:14Antipatterns
- 36:19Bill
- 37:49Outro
related talks
39mThe Ticking Timebomb of Observability Expectations
18mHow not to practice observability
22mObservability is Too Expensive!
26mDriving O11y x to Nx: Probing Bullet Holes and not Breaking the Bank
32mEssential Software Observability for Every Role: Look No Further
27mOpening the Door to Observability
50mPanel: Observability
10mLightning Talk: Observability Debt: When Telemetry Stops Telling the Truth