Robust Energy Management for Green and Survivable IP Networks

Abstract: Despite the growing necessity to make Internet greener, it is worth pointing out that energy-aware strategies to minimize network energy consumption must not undermine the normal network operation. In particular, two very important issues that may limit the application of green networking techniques concern, respectively, network survivability, i.e. the network capability to react to device failures, and robustness to traffic variations. We propose novel modelling techniques to minimize the daily energy consumption of IP networks, while explicitly guaranteeing, in addition to typical QoS requirements, both network survivability and robustness to traffic variations. The impact of such limitations on final network consumption is exhaustively investigated. Daily traffic variations are modelled by dividing a single day into multiple time intervals (multi-period problem), and network consumption is reduced by putting to sleep idle line cards and chassis. To preserve network resiliency we consider two different protection schemes, i.e. dedicated and shared protection, according to which a backup path is assigned to each demand and a certain amount of spare capacity has to be available on each link. Robustness to traffic variations is provided by means of a specific modelling framework that allows to tune the conservatism degree of the solutions and to take into account load variations of different magnitude. Furthermore, we impose some inter-period constraints necessary to guarantee network stability and preserve the device lifetime. Both exact and heuristic methods are proposed. Experimentations carried out with realistic networks operated with flow-based routing protocols (i.e. MPLS) show that significant savings, up to 30%, can be achieved also when both survivability and robustness are fully guaranteed.