The Science of Signals: Mastering Telemetry for Observability
summary
Covers concrete observability choices: uses the OpenTelemetry Java SDK and a local collector + Java agent for auto‑instrumentation, sends OTLP (binary) instead of JSON (traffic drops ~10×), and builds a queuing pipeline/collector and custom exporter that works in production. Explains metrics vs logs vs traces vs profiling, why head vs tail sampling matters, when to prefer histograms/exponential histograms and exemplars, and warns about 10,000 cardinality limit per metric; describes dropping ~70% of telemetry and reaggregating to cut cost and make alerts reliable.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 0:26Speaker introductions
- 3:00Talk agenda overview
- 4:04History of metrics
- 10:03Evolution of logs
- 14:56Introduction to traces
- 19:07Man and the machine dashboards
- 22:39Error reporting challenges
- 25:13Alerting and notifications
- 40:33Signals deep dive metrics
- 1:11:29Metric backends and storage
- 1:27:44Break time
- 1:48:26Logs and structured logging
- 2:02:02Traces and distributed tracing
- 2:16:13OpenTelemetry overview
- 2:25:29OpenTelemetry demo
- 2:31:37Demo walkthrough
- 2:40:23Q&A and workshop reminder
related talks
36mObservability For You and Me with OpenTelemetry
27mThe Signal in the Storm: Practical Strategies for Managing Telemetry Overload
52mHow to effectively spy on your systems
20mObservability and Open Telemetry
37mObservability For You And Me With Opentelemetry
32mOpenTelemetry feat. logs, metrics and traces
50mGuide to Observability with OpenTelemetry
22mShift Into an Observability Mindset with OpenTelemetry