The Structure and Interpretation of Graphs
summary
Shows concrete graph failures and fixes: latency percentiles (50/85/95/99) on a log scale reveal huge tail impacts and timeouts; QPS hourly spikes expose client-side sync; a hash of empty string plot detects missing input; serial-number sorting of disk failures exposes bad batches and reserialised repairs; load‑balancer change narrows latency distribution but raises the floor; client underdamped retries replace exponential backoff. Recommends totals, log scales, distributions, multi‑resolution dashboards and tools like Grafana/Prometheus/Graphite.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 7:11Graphs are Obvious
- 12:00Graphs are Less Obvious
- 18:10Storage Right Failure Events
- 21:44Retry Behavior
- 27:12Recommendations
- 30:08Better visualizations
- 30:40Population Pyramid
- 33:18Sinkey Diagrams
- 35:00Heat Map
- 35:53Electoral College
related talks
06mPerformance is about more than time series charts
45mA Practical Guide to Dashboarding
21mGraphing tools for scheduler tracing
26mImproving How We Observe Our Observability Data: Techniques for SREs
49mA Practical Guide to Dashboarding
27mCharts and visual lies
10mTurning data into leverage
26mImproving Observability in Your Observability: Simple Tips for SREs