Lightning Talk: Running Energy-Aware AI Inference on Edge Kubernetes With Kepler
Lightningat Kubernetes on Edge Day North America 2025 · on Monday, 2025-11-10· at 10:00· Building B | Level 2 | B203·
summary
Deploys a lightweight K3s + Kepler + Prometheus stack on a laptop-emulated edge cluster and runs an ONNX Runtime image classifier while exporting power metrics. Measurements show a baseline of ~2.2–2.3 joules and inference spikes that exceed baseline by about 5x, demonstrating real-time visibility without GPUs or extra hardware. The demo is repeatable, uses only two core open-source tools, and the setup successfully correlates inference activity to measurable energy impact.
AI-written summary · about
related talks
07mLightning Talk: AI at the Edge: ONNX Inference in WASM on Featherweight K0s
27mGreening the AI Cloud: Validating Power Models for Kubernetes Containers
35mSustainability Research the Cloud Native Way
40mSustainability the Container Native Way
26mBuilding a Cloud Native Edge Platform for Accelerating AI Inference
32mKeynote: Powering Generative AI with Kubernetes
12mLightning Talk: Benchmarking Your Distributed ML Training on the K8s Platform
54mEconomics and Best Practices of Running AI/ML Workloads on Kubernetes