Tracing, Fast and Slow: Digging into and Improving Your Web Service's
summary
Practical checklist for tracing a web service: instrument with metadata propagation (Dapper-style) or fall back to simple UUID headers and log correlation; expect tracing to add measurable overhead (Google reports ~1.5x throughput hit and ~16% higher latency when tracing everything), so sample between 0.01% and 10% and choose head-based sampling for steady percentiles and tail-based for 99th-percentile anomalies. Tools that actually ship: Zipkin, Jaeger, OpenTracing/OpenCensus, Stackdriver, AWS X‑Ray; use request flow graphs and calling-context trees to find the bottleneck.
AI-written summary · about
chapters (9)
- 0:00Introduction to tracing
- 1:37Why we need distributed tracing
- 3:20Tracing and distributed systems
- 7:56Three common tracing approaches
- 12:53Best practices for instrumentation
- 18:56Sampling strategies
- 20:37Visualization techniques
- 22:26Troubleshooting performance issues
- 26:26Tools and industry standards
related talks
31mUnderstanding Microservices with Distributed Tracing
50mDistributed Tracing in the Wild
16mPragmatic execution tracing
42mDistributed Tracing Opening the Door to Visualisation and Correlation
34mWould You Like Some Tracing With Your Monitoring?
22mDistributed Tracing
26mDistributed Tracing FAQ
59mAn introduction to distributed tracing