Energy-Efficiency in Optical Networks

Abstract

This thesis expands the state-of-the-art on the complex problem of implementing energy efficient optical networks. The main contribution of this Ph.D. thesis is providing a holistic approach in a multi-layered manner where different tools are used to tackle the urgent need of both estimating and optimizing power consumption in different network segments. An energy consumption analysis for a novel digital signal processing for signal slicing to reduce bandwidth requirements for passive optical networks is presented in this thesis. This scheme aims at re-using low bandwidth equipment to cope with current traffic demands and this dissertation tackles the trade-off between energy efficiency and quality of service in terms of latency. Another important contribution of this thesis is the novel mixed integer linear programing (MILP) formulation for internet protocol (IP) over wavelength division multiplexing (WDM) core network design. This work provides a detailed model based on a modular architecture where the network is upgraded as the network traffic increases considering physical constraints instead of approximations on power consumption per port or assuming a single type of interface module. A comprehensive analysis on the trade-off between power consumption and capital expenditures (CAPEX) is also presented. The results confirmed that gross-grained designs (i.e., designs that account for high bandwidth technologies) are attractive from a cost perspective, however, power consumption needs to be further improved. A second analysis for core networks is performed on a programmable networking platform, named software defined networking, to consolidate core network designs towards cost-efficient and flexible solutions. Results on jitter, latency, and power consumption for a novel south-bound protocol named KeyFlow are reported showing more than 50% reduction in the round trip time, eliminating jitter from the system, and obtaining up to 57% power reduction compared to a reference OpenFlow switch. Furthermore, this thesis introduces an evaluation of a novel approach for short range high capacity links for datacenters. A polarization multiplexing strategy named quad-polarization, where four independent data streams are transmitted simultaneously, is evaluated from computational complexity, power consumption, and receiver sensitivity perspectives. To provide a comprehensive analysis, comparison with parallel optics and WDM systems is reported. These results show the trade-off between increased capacity and both power consumption and system performance. In conclusion, an energy-efficient set of tools has been provided covering different aspects of the telecommunication network resulting in a cohesive research in three networks segments: access, core networks, and datacenters. Additionally, it opens prospects for next generation energy efficient networks providing tools to analyze, estimate, and optimize that new metric in optical network design: energy consumption.