🤖 AI Summary
In virtualized environments, guest VMs lack physical power interfaces and host privileges, hindering direct energy consumption measurement and impeding energy-aware scheduling and cost optimization. This paper proposes a purely client-side virtual server power consumption prediction method that estimates energy usage solely from guest-observable resource utilization metrics—CPU, memory, disk I/O, and network traffic—using a gradient boosting regression model, without requiring host access or privileged instrumentation. Ground-truth power measurements are obtained via RAPL on the physical host for model training and validation. Experimental evaluation across diverse representative workloads achieves R² scores of 0.90–0.97, demonstrating, for the first time, the feasibility of high-accuracy virtual server power estimation using only guest-side resource data. This work fills a critical gap in cloud computing by enabling non-intrusive, low-privilege energy modeling.
📝 Abstract
This paper presents a machine learning-based approach to estimate the energy consumption of virtual servers without access to physical power measurement interfaces. Using resource utilization metrics collected from guest virtual machines, we train a Gradient Boosting Regressor to predict energy consumption measured via RAPL on the host. We demonstrate, for the first time, guest-only resource-based energy estimation without privileged host access with experiments across diverse workloads, achieving high predictive accuracy and variance explained ($0.90 leq R^2 leq 0.97$), indicating the feasibility of guest-side energy estimation. This approach can enable energy-aware scheduling, cost optimization and physical host independent energy estimates in virtualized environments. Our approach addresses a critical gap in virtualized environments (e.g. cloud) where direct energy measurement is infeasible.