Energy-Aware Adaptive Routing Solutions in IP-over-WDM Networks

Abstract

Today’s core networks are permanently powered on and consume non-negligible amount of energy. Traffic varies over time giving an opportunity to switch off or put into standby mode a subset of network devices in order to save energy in low-demand hours. Line cards are targeted to be dynamically switched on and off since their activation and deactivation times are expected to be sufficiently small, their power consumption is significant (400–500 W per line card), and a sufficiently large set of line cards is usually installed in the network. The focus of this thesis is on EnergyAware Adaptive Routing Solutions (EA-ARSs) which utilize traffic variation, traffic rerouting and sleep modes in order to save energy consumed by line cards in Internet Protocol (IP)-over-Wavelength Division Multiplexing (WDM) networks. We collect from publicly available sources (product data sheets, research articles, and databases) extensive sets of realistic input data for the EA-ARSs that are crucial for energy saving, namely traffic data and power consumption data of single network devices. Extensive sets of traffic matrices containing traffic demands between all node pairs in the network cover periods of days, weeks, and months with granularity of 5 min, 15 min, a day, and a month. Physical supply topologies corresponding to the traffic matrices are reported too. The power consumption of single network devices determines the amount of power that can be saved in the whole network by switching off subsets of devices. Using the realistic input data we estimate the potential power and energy saving assuming different levels of freedom of rerouting in an IP-over-WDM network. The proposed approaches Fixed Upper Fixed Lower (FUFL), Dynamic Upper Fixed Lower (DUFL) and Dynamic Upper Dynamic Lower (DUDL) differ with the possibility of rerouting of traffic demands over the logical (IP) topology, and the possibility of rerouting of lightpaths over the physical topology. A Static Base Network (SBN) is used as a starting point for EA-ARSs. The SBN determines devices installed in the network. Sophisticated Mixed-Integer Linear Programming (MILP) formulations are used for the SBN design as well as for DUFL and DUDL computations, which are highly complex optimization problems. FUFL is a simple, fully distributed heuristic. Our results show that flexibility of the IP routing (Dynamic Upper) contributes the most to the energy saving. Additional flexibility of routing in the WDM layer (Dynamic Lower) brings marginal savings. The simple approach FUFL brings significant savings, which are not as high as the savings brought by DUFL though. Furthermore, they depend on the ratio of the maximum total traffic demand and the capacity of a WDM channel. Spatial distribution of traffic demands also influences the energy savings achieved with FUFL. A set of evaluation criteria for EA-ARSs is proposed in order to determine the approaches which have a chance to be implemented in an operational network of the future. In this context, an adaptive heuristic called Energy Watermark Algorithm (EWA) is proposed. EWA uses network configuration from previous time period in order to calculate new network configuration. This speeds up calculation of network configuration and reduces reconfiguration costs. Furthermore, EWA has a set of parameters, which allows a network operator to find the preferred trade-off between energy