Evolution of on-Node Adaptive Power Tuning

This talk discusses strategies for optimizing on-node power consumption to reduce costs and improve sustainability. The speakers highlight the limitations of current static power-tuning methods and propose a more adaptive approach using TuneD and the Kubernetes Power Manager. They delve into CPU capabilities like P-states (frequency scaling) and C-states (sleep states), and how these can be configured and controlled within a Kubernetes environment through tools like the Node Tuning Operator and the Intel Kubernetes Power Manager. The presentation includes a demonstration of applying these power management techniques to microservice workloads.