In this paper we present a novel experimental platform for network management able to dynamically optimize the energy consumption of backbone IP networks operating with OSPF. The idea is to efficiently adjust link weights to put to sleep idle devices. The framework relies on multiple pre-computed sets of link weights that are applied to the considered network domain according to real time measurements on the link utilization. The management module exploits the Simple Network Management Protocol (SNMP) to collect the load measurements and modify the link weights. The pre-computed link weights are calculated by running a state-of-the-art algorithm for off-line energy-aware traffic engineering based on predicted traffic matrices. The modules of the platforms have been implemented for Linux based environments and tested using emulated networks of virtual machines. Experimental results show that up to 50% of energy savings can be achieved during low traffic periods without compromising network performance and stability.
Paru en mai 2013 , 18 pages