How Green Is My OpenTelemetry Collector?
Talkat KubeCon + CloudNativeCon Europe 2025 · on Wednesday, 2025-04-02· at 17:45· Level 1 | Hall Entrance N10 | Room E·
summary
Kepler exports a metric called kepler container_joules_total which the demo converts to kWh to measure collector energy use; experiments compare a single contrib collector versus splitting into two pipeline-specific collectors and versus a custom collector distro built with only needed components. Using the OTel Operator (target allocator + Prometheus receiver), kubelet stats and Prometheus, results show splitting into two instances raises total energy (worse), while the custom distro often lowers memory and sometimes lowers energy but with variable runs, so measure, change one variable, and repeat.
AI-written summary · about
chapters (18)
- 0:00Welcome
- 1:25IT emissions impact
- 3:13OpenTelemetry's carbon role
- 3:55Introducing Kepler
- 6:46Kepler architecture
- 8:30Observability stack setup
- 10:16OpenTelemetry operator config
- 13:45Collector pipeline design
- 14:48Collector tuning experiments
- 15:52Energy and memory measurement
- 17:15Pipeline split results
- 18:18Custom collector build
- 19:00Key learnings
- 21:48Next steps
- 22:29Green optimization strategies
- 23:35Expert insights
- 25:44Resources and community
- 27:10Closing
related talks
17mKepler: Powering up the efficiency of your cluster
25mOpenTelemetry Collector Deployment Patterns
40mSustainability the Container Native Way
57mTelemetry Showdown: Fluent Bit vs. OpenTelemetry Collector - A Comprehensive Benchmark Analysis
23mControlling Data Overhead with the OpenTelemetry Collector
16mUsing the OpenTelemetry Collector to Empower End Users
40mMeasuring Energy Consumption in Clusters with Kepler
24mTelemetry Showdown: Fluent Bit Vs. OpenTelemetry Collector - A Comprehensive Benchma...