Physical Layer Aware Optical Networks

This thesis describes novel contributions in the field of physical layer aware optical networks. IP traffic increase and revenue compression in the Telecom industry is putting a lot of pressure on the optical community to develop novel solutions that must both increase total capacity while being cost effective. This requirement is pushing operators towards network disaggregation, where optical network infrastructure is built by mix and match different physical layer technologies from different vendors. In such a novel context, every equipment and transmission technique at the physical layer impacts the overall network behavior. Hence, methods giving quantitative evaluations of individual merit of physical layer equipment at network level are a firm request during network design phases as well as during network lifetime. Therefore, physical layer awareness in network design and operation is fundamental to fairly assess the potentialities, and exploit the capabilities of different technologies. From this perspective, propagation impairments modeling is essential. In this work propagation impairments in transparent optical networks are summarized, with a special focus on nonlinear effects. The Gaussian Noise model is reviewed, then extended for wideband scenarios. To do so, the impact of polarization mode dispersion on nonlinear interference (NLI) generation is assessed for the first time through simulation, showing its negligible impact on NLI generation. Thanks to this result, the Gaussian Noise model is generalized to assess the impact of space and frequency amplitude variations along the fiber, mainly due to stimulated Raman scattering, on NLI generation. The proposed Generalized GN (GGN) model is experimentally validated on a setup with commercial linecards, compared with other modeling options, and an example of application is shown. Then, network-level power optimization strategies are discussed, and the Locally Optimization Global Optimization (LOGO) approach reviewed. After that, a novel framework of analysis for optical networks that leverages detailed propagation impairment modeling called the Statistical Network Assessment Process (SNAP) is presented. SNAP is motivated by the need of having a general framework to assess the impact of different physical layer technologies on network performance, without relying on rigid optimization approaches, that are not well-suited for technology comparison. Several examples of applications of SNAP are given, including comparisons of transceivers, amplifiers and node technologies. SNAP is also used to highlight topological bottlenecks in progressively loaded network scenarios and to derive possible solutions for them. The final work presented in this thesis is related to the implementation of a vendor agnostic quality of transmission estimator for multi-vendor optical networks developed in the context of the Physical Simulation Environment group of the Telecom Infra Project. The implementation of a module based on the GN model is briefly described, then results of a multi-vendor experimental validation performed in collaboration with Microsoft are shown.