Investigating Performance Issues In Microservices Arch.
summary
A multi-tenant SaaS running on Kubernetes and Docker across AWS and Azure instruments logs and metrics but still faces sporadic UI slowdowns; adopting Jaeger and OpenTelemetry yields concrete results. Traces expose a staircase pattern of needless serial calls that convert to concurrent calls and cut latency, and reveal a missing cache causing per-request times to swing from 100 ms to 10 s. Auto-instrumentation across Node.js, Java and DB layers produces live system diagrams and pinpoints the longest spans to fix.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 1:17Distributed tracing for performance
- 2:23Observability journey challenges
- 3:06Logs and metrics limitations
- 5:11Request ID propagation attempt
- 7:38Why tracing became essential
- 9:46How tracing works visually
- 10:30Choosing Jaeger
- 12:40Instrumenting apps for tracing
- 15:11Serial call sequences discovery
- 16:36Caching and DB latency discovery
- 18:45Live system diagrams
- 19:27Adopting tracing organization-wide
- 20:08Tips for tracing start
- 21:13OpenTelemetry future
- 22:38Auto instrumentation benefits
- 24:03Filtering traces by duration
- 25:06Correlating traces with logs
- 25:48Trace pattern analysis
- 30:25Summary of tracing benefits
- 31:06Q&A
related talks
40mCracking Microservices Performance Issues with Distributed Tracing
14mLeveraging Distributed Tracing in Kubernetes & Containerized Environments
31mUnderstanding Microservices with Distributed Tracing
52mIs your JVM app flying blind? Unmask hidden issues with Observability superpowers!
19mOops my app performance sucks... what if observability could be a game changer?
35mJaeger Intro
41mPractical introduction to OpenTelemetry tracing
45mObservability for Modern Spring Applications