Performance and Energy Efficiency in Docker
summary
An automated CI pipeline measures per-container energy and performance by mapping PIDs to containers and correlating power with API metrics, using Locust to generate load, Scaphandre to sample per-process power, and cAdvisor plus a GitLab runner to collect OS metrics and deploy via Docker Swarm. The system runs on one test machine, exposes recurring latency hiccups whose 95th‑percentile spikes coincide with power surges, and shows initial requests burn significantly more energy; Kubernetes has not been tested.
AI-written summary · about
chapters (8)
- 0:00Introduction to talk
- 1:05Why performance and power matter
- 2:33Tools for testing and measurement
- 4:43Automated testing pipeline
- 6:31Measuring container power usage
- 8:15Analyzing performance hiccups
- 9:42Automated testing benefits
- 10:08Q&A session
related talks
22mEmbedded Software Development Efficiency Improvement with Docker and Kubernetes
08mTowards Greener Deployment: Assessing Energy Efficiency i...
30mArchitecting High-Density Performance Testing with Container-Based Simulation
20mMonitoring Energy Consumption and Building Energy Efficient Systems the Cloud Native Way
23mFull-stack energy consumption test bench for developers
35mSustainability Research the Cloud Native Way
43mLinux Container Performance Analysis
01h 18mEmbracing Docker simplicity - while harnessing platform power