The open-source SusQL Operator (susql.org) enables administrators of Kubernetes clusters such as Red Hat's OpenShift to monitor and visualize the total energy consumption and carbon emission for multiple long-running user-specified workloads. For example, SusQL could track all of the various jobs involved in training an AI model over a period of days, weeks, months, or longer. This helps developers of AI models comply with regulations to report the energy consumption of specific AI model training. SusQL can flexibly target any user-specified grouping including by projects, teams, organizations, clients, etc. by utilizing the newly introduced LabelGroup annotation.

As a lightning talk this will be a brief introduction to the concepts, functionality, and status of the SusQL project. The speaker looks forward to more detailed discussions in the networking time that follows.