Why Large-Scale Observability Needs Graph
summary
Argument: observability lacks the system model needed to infer internal state, so connect metrics/logs/traces to an explicit graph model. Production graph tech cited: Jaeger service graph (visual-only trace topology), graph DBs neo4j AWS Neptune Azure Cosmos, and tooling examples like Backstage service catalog (YAML-driven central catalog) and eBay's Groot for graph-based root-cause analysis. Kiali overlays Prometheus on Istio service maps. Proposal: store a property/knowledge graph across tools and consider an open “observability graph” standard.
AI-written summary · about
chapters (12)
- 0:00Intro
- 1:08Observability challenges
- 2:27We've been missing something...
- 5:29Control system observability
- 6:29Software system observability Where is the problem?
- 7:38Small-scale application
- 8:44Large-scale application(s)
- 9:48The model is a graph
- 10:47Graph 101
- 15:53Can graph be used to model our systems?
- 21:25Extending our scope of observability
- 24:55Do we need an open standard for the "observability graph"?
related talks
39mThe 5W’s of Observability with Practical Examples
16mThe State of Observability
13mObservability Where are We Headed?
36mObservability? It's a Data Analytics Problem, You Fool!
43mObservability Visualization in the Age of OpenTelemetry
30mObservability is More than Logs, Metrics & Traces
32mUnderstanding Observability in Kubernetes
13mArchitecting an open observability stack