The paper presents a dynamic resource provisioning framework for virtualized server environments, aiming to reduce power consumption and improve energy efficiency while maintaining quality-of-service (QoS). The authors propose a lookahead control (LLC) scheme that accounts for switching costs and encodes risk in the optimization problem. Experiments using IBM's Trade6 application demonstrate that the proposed approach saves an average of 22% in power consumption over a 24-hour period compared to a system without dynamic control, while maintaining QoS goals. The controller's performance is further analyzed using trace-based simulations for larger server clusters, and the use of neural networks is explored to reduce computational overhead. The results show that a risk-aware controller with a risk preference factor of 2 reduces both SLA violations and host switching activity, leading to better performance.The paper presents a dynamic resource provisioning framework for virtualized server environments, aiming to reduce power consumption and improve energy efficiency while maintaining quality-of-service (QoS). The authors propose a lookahead control (LLC) scheme that accounts for switching costs and encodes risk in the optimization problem. Experiments using IBM's Trade6 application demonstrate that the proposed approach saves an average of 22% in power consumption over a 24-hour period compared to a system without dynamic control, while maintaining QoS goals. The controller's performance is further analyzed using trace-based simulations for larger server clusters, and the use of neural networks is explored to reduce computational overhead. The results show that a risk-aware controller with a risk preference factor of 2 reduces both SLA violations and host switching activity, leading to better performance.