How to (easily) evaluate the energy consumption of your software ?
summary
Shows a reproducible test bench: a laptop running the voice API in Docker, stress tests via REST-assured, and either a physical power meter or PowerAPI (INRIA open source) as a software sensor. Measures show idle vs load: machine draws about 48 W at peak with a ~20–28 W increase under load, performance flattens around 6 calls/sec and errors rise after 8 calls/sec, and per-call energy reaches an observed optimum near 5 joules per call. Telephony servers account for ~80% of consumption; plan is CI integration and sizing/bottleneck fixes.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:15Welcome
- 3:00What is Fast for Voice
- 4:05The Sustainable Challenge
- 5:30Impact by Design Approach
- 7:30Measure
- 8:45Test Bench
- 10:55Demo
- 15:40Ideas
- 17:00Conclusion
- 18:15Questions
related talks
22mHow to (easily) evaluate the energy consumption of your software ?
42mEnergy-Efficient Software Architecture for Developers
50mEnergy Efficient Design Principles of Software Engineering
09mAplikacionet e gjelbra - efikasiteti i energjisë në software
13mPerformance and Energy Efficiency in Docker
31mCreate an eco-friendly world with Green Software Engineering
39mCreate an Eco-Friendly World with Green Software Engineering
49mThe Future is 💚: A Journey through the Energy Efficient ..