Abstract
Reducing energy consumption in the Telecom industry has become a major research challenge to the Internet community due to high level of energy waste on redundant network devices. In search for a paradigm shift, recent research efforts have been focusing on time-driven sleep-mode reconfiguration of network elements during periods of low traffic demand. However, due to the routing re-convergence issue of today's traditional IP routing protocols, frequent network reconfigurations are generally deemed to be harmful as a result of routing table re-convergence. Furthermore, diurnal traffic behaviours are unpredictable and can lead to network congestion as a result of the reduced network resources. This thesis presents novel event-driven green backbone routing schemes for network managements which are capable of saving energy in fixed IP networks (using both regular and non-regular traffic matrix) without inhibiting its performance. First, a Link Wake-up Optimisation Technique (LiWOT) is proposed during energy saving periods when the pruned topology is applied. The key novelty here is that LiWOT selects the minimum number of router's line cards to wake-up when the network is congestion is detected. This is contrary to the norms of reverting to the full network topology or on-the-fly network reconfigurations in the case of even minor traffic surge and thereby sacrificing energy savings. In order to mitigate the effect of routing re-convergence in networks, LiWOT prioritises the waking up of non-disruptive sleeping links. This scheme was further extended to a fully disruption-free scheme. The second proposed scheme is the Green Link Weight Disruption-Free Energy-aware Traffic Engineering which limits its wake-up operation to only non-disruptive links. In order to maximise the energy savings, the number of this type of links are maximised in an offline manner. Using a genetic algorithm based approach, a new link weight optimisation scheme is proposed and this forms the basis of the second research contribution. Finally, a completely dynamic link sleeping reconfigurations (DLSR) for green traffic engineering is proposed. The scheme coordinates the sleep and wake-up operations in a dynamic way such that operations are based on the current traffic. The key contribution is that DLSR is oblivious of historical traffic conditions like the previous schemes and can enhance energy savings by putting back woken-up links to sleeping mode during low traffic. The performances of the three schemes were evaluated using the publicly accessible traffic traces of both GEANT and Abilene network respectively over a period of one week and the obtained results show a substantial amount of energy saving.