Observability for Emerging Infra: What Got You Here Won’t Get You There
summary
Argues that observability ≠ monitoring and proves it with Parse’s rescue using Facebook’s Scuba: shipping raw event datasets let engineers slice by user ID and debug unknown unknowns that metrics missed. Recommends instrumenting every network hop, storing sampled structured events (high-cardinality), avoiding pre-aggregating read-time metrics, and treating dashboards as disposable — dashboards are technical debt. Claims this lets teams catch ~80% of regressions in prod, reduce noisy paging to errors/latency/request-rate, and restore true software ownership.
AI-written summary · about
chapters (8)
- 0:00The evolution of ownership
- 3:52Defining observability
- 7:44Monitoring versus observability
- 14:06Distributed system challenges
- 20:44The case for exploration
- 26:55Required tooling capabilities
- 33:32Replacing dashboards with data
- 41:46Reducing alert fatigue
related talks
43mObservability for Emerging Infra: What Got You Here Won't Get You There
39mObservability for Emerging Infra: What Got You Here Won't Get You There
46mObservability for Emerging Infra: What Got You Here Won't Get You There
52mObservability: it's not just an ops thing
29mObservability for Developers: How to Get From Here to There
29mObservability; a gentle introduction
31mObservability for Developers: How to Get There
13mObservability Where are We Headed?