Currently, there is no solution available to accurately identify the process or processes consuming the most energy within a Virtual Machine (VM) or, in the case of a service provider, a Virtual Network Function (VNF).

The energy-intensive process within the workload can vary significantly depending on the VNF’s function, whether it is routing, firewalling, load balancing, VPN, or other network services.

The key challenge is to enable the hypervisor to provide the necessary registers for power-monitoring tools, such as Kepler, within the VM—without requiring any modifications to the Guest OS or reliance on machine learning techniques.

Technical documentation: https://docs.google.com/document/d/1OIibA7Ti0CFr3c8ZS2etdSw1eyjeorEXdLmM5_WWOzQ/edit?usp=sharing