Cloud infrastructure emits up to 4 % of the world's carbon emissions. With the trend of shifting more and more workloads to cloud infrastructure being able to manage this through the lens of not only economical but also ecological reasoning is a must.

Creating accurate energy models for cloud infrastructure is a complex but essential step toward achieving sustainable digital systems. In the Eco:Digit project, we embarked on a journey to measure and understand energy consumption in virtualized cloud environments, uncovering the challenges of correlating software workloads with physical resource usage.

This talk shares the process of developing energy profiles for cloud infrastructure, focusing on the use of open-source conform SCS clouds to create an energy model. We will discuss the technical and methodological hurdles we faced—such as accessing host-level metrics in virtualized environments, calibrating virtual consumption against physical measurements facing heterogeneous hardware, and dealing with limited transparency in hyperscaler clouds.

Key lessons shared include how to design a test environment that reliably maps energy consumption across physical and virtual layers, the importance of integrating diverse monitoring tools, and the role of collaborative partnerships with cloud providers like PlusServer and ScaleUp. Finally, we'll present the current state of our energy profiling system, showcasing its capabilities for generating detailed energy consumption reports and enabling actionable sustainability insights. This talk will provide the audience with technical takeaways and practical knowledge for anyone interested in energy monitoring, open-source cloud infrastructure, or advancing sustainability in the digital domain.