Monitoring at Scale: Intuitive Dashboard Design
summary
At about 10,000 messages per second the claim is that dashboards must let System One spot normality and only flip attention when something deviates; concrete tools cited are StatsD and Graphite with Xenos/Cacti plugins and a PHP weather map for topology views. Practical rules: show 50th/95th percentiles not averages, normalize scales or use derivatives to flatten periodic patterns, color heatmaps of services and links expose bottlenecks with minimal attention.
AI-written summary · about
chapters (14)
- 0:00Intro
- 3:10Cognitive Science
- 6:15Fast Thinking
- 8:42Infographics
- 10:36dashboards are not analysis tools
- 11:41fatigue in our brain
- 14:37information needs to be organized
- 16:07Embellishments
- 18:28The Brain
- 21:51Bias
- 26:59Instrumentation
- 28:54Distribution
- 36:59Heat maps
- 41:38Questions
related talks
49mA Practical Guide to Dashboarding
41mMetrics Applied
46mHow many dashboards does your Grafana have?
37mUX Design and Education for Effective Monitoring Tools
57mHow to Display 100,500 Metrics of a Distributed System and Not Go Insane
22mService Monitoring Manual—2018 Edition
54mA Practical Guide to Dashboarding
26mImproving How We Observe Our Observability Data: Techniques for SREs