During peak hours, the focus of an Internet service provider (ISP) is on meeting user demand and maintaining the quality of service of a network. However this leads to under-utilization of network devices/components in different parts of the network. Since the power consumption of network devices are constant and independent of their actual workloads, the under-utilized network devices/components continue to consume the same power as that of a component working at full workload. We propose an energy aware network management system (NMS) framework based on the centralized decision management system (CDMS) reported in [1]. The system uses Bayesian belief network (BBN) to predict the utilization in different ports of network routers. The information already available in the NMS is sufficient for the proposed framework. For implementing the proposed scheme, an extended interface management information base (MIB) is proposed, which includes an operational state. Simulations are performed on two standard networks - IndiaNet and ARPANet, and the results show that with the proposed approach the CDMS is able to save approximately 10-16% of power at the cost of 1-5% end-to-end delay overhead. The cost of re-routing and Simple Network Management Protocol (SNMP) messages is minimal and acceptable considering the total power saved by CDMS. © 2012 IEEE.