Power Measurement and Attribution in Kernel for Embedded System Processes
summary
Kernel-side framework claims per-process energy attribution by combining multivariate regression models with device-specific calibration sweeps (idle baseline, on/off and brightness sweeps) and process inputs from proc accounting so energy is apportioned by runtime fraction. Prototype uses power probes, battery/current sensors or Powertop/NVIDIA SMI–style counters to collect parameters; example anomalies include a reported 1.45 W for GNOME Shell and CPU examples (15 W over 5 ms) showing power is non‑linear. Output is an estimated, not ground‑truth, CLI/IDE API; accuracy, bias tradeoffs and privacy remain open.
AI-written summary · about
chapters (8)
- 0:00Power measurement challenges
- 7:32Energy measurement importance
- 13:56Energy attribution framework
- 18:55Model implementation challenges
- 23:37Energy and carbon footprint
- 24:59Practical applications
- 27:06Key takeaways
- 28:10Q&A
related talks
39mAbout The Need to Power Instrument The Linux Kernel
42meBPF-Powered Observab...
17mKepler: Powering up the efficiency of your cluster
51mLinux Systems Performance - Architecture
40mSustainability the Container Native Way
44mThink Outside the VM: Unobtrusive Performance Measurement
29mAuto-instrumentation for GPU performance using eBPF
21mAuto-Instrumentation for GPU Performance using eBPF