Greening the AI Cloud - Validating Power Models for Kubernetes Containers

Explore the critical issue of energy consumption in Machine Learning (ML) and its impact on data centers in this 27-minute talk by Marcelo Amaral from IBM Research Tokyo. Learn about the Kepler framework, which estimates power consumption at various levels in Kubernetes environments. Discover how to create and validate power models for optimizing energy usage in ML workloads, addressing the growing need for sustainable AI practices and reduced greenhouse gas emissions. Gain insights into the observability of ML power consumption and its importance in data center management, especially with the rising popularity of Foundation Model systems.