Improving How We Observe Our Observability Data: Techniques for SREs
summary
Argues that clearer observability comes from combining related metrics into single visuals rather than dumping dozens of separate charts, demonstrating a reusable pattern that plots four different metrics (deployments, per-second volume by status code as bars, latency as a line, and time) and recommends sparklines (brew install spark) to pack hundreds more data points into tables. Shows historical examples (Playfair, Minard), warns about skewed scales, shading, missing keys, truncated axes, and swapping averages for p50 p75 p90 p95 and using orthogonal axes for different cardinalities.
AI-written summary · about
chapters (8)
- 0:00Intro
- 1:13What we're going to talk about
- 2:23Time Series Data Motion of the Planets
- 9:29Multivariate Time Series Chart Steal Playfair's chart!
- 10:52Multivariate Observability Visualizations The Minard
- 16:32Common pitfalls that distract from good observability
- 22:53Improving Data Density and Observations in Tables Sparklines
- 24:30Conclusion
related talks
26mImproving Observability in Your Observability: Simple Tips for SREs
29mObservability for Developers: How to Get From Here to There
29mObservability; a gentle introduction
32mEssential Software Observability for Every Role: Look No Further
40mDemystifying Observability and Monitoring
32mUnderstanding Observability in Kubernetes
31mNo-Nonsense Observability Improvement
31mObservability for Developers: How to Get There