<p dir="ltr">Recent advancements in large language models such as ChatGPT and AU Large allow for the effective integration and application of LLMs into network devices such as switches and routers in terms of the ability to play a role in configuration and management. The given devices are an essential part of every network infrastructure, and the nature of physical networking topologies is complex, which leads to the need to ensure optimal network efficiency and security via meticulous and precise configurations.</p><p dir="ltr">The research explores the potential of an AI-driven interface that utilizes AU Large to streamline, enhance, and automate the configuration process of network devices while ensuring that the security of the whole process is guaranteed by running the entire system on-premise. Three core areas are of primary concern in the given study: the effectiveness of integrating the AU Large into network management systems, the impact on efficiency, accuracy, and error rates in network configurations, and the scalability and adaptability to more complex requirements and growing network environments.</p><p dir="ltr">The key performance metrics evaluated are the error rate in the generated configurations, scalability by looking at the performance as more network devices are added, and the ability to generate incredibly complex configurations accurately. The high-level results of the critical performance metrics show an evident correlation between increased device count and increased prompt complexity with a degradation in the performance of the AU Large model from Mistral AI.</p><p dir="ltr">This research has significant potential to alter preset network management practices by applying AI to make network configuration more efficient, reduce the scope for human error, and create an adaptable tool for diverse and complex networking environments. This research contributes to both AI and network management fields by highlighting a path toward the “future of network management.”</p>

Optical communications systems operate at the limits of their margins to respond to increasing capacity demands. Some of the signal processing functions required must soon operate at speeds beyond electronic implementation. Optical signal processors are fundamentally analog which requires precise control of the operating state. Programmable optical components are consequently essential. The thesis explores and elucidates the properties of meshes of generalized Mach-Zehnder interferometers (GMZIs) amenable to silicon (Si) photonics integration that are based on multimode interference couplers with programmability achieved via voltage controlled phase-shift elements within the interferometer arms to perform a variety of finite impulse response (FIR) and infinite impulse response (IIR) signal processing functions. The thesis presents a novel class of integrated photonic phased array systems with a single-stage, multistage, and feedback architectures. The designed photonic integrated systems utilize nano-electromechanical-system (NEMS) operated phase shifters of cascaded free suspended slot waveguides that are compact and require a small amount of power to operate. The structure of the integrated photonic phased array switch (IPPAS) elements is organized such that it brings the NEMS-operated phase shifters to the exterior sides of the construction; facilitating electrical connection. The transition slot couplers used to interconnect the phase shifters to the rest of the silicon structure are designed to enable biasing one of the silicon beams of each phase shifter from an electrode located at the side of the phase shifter. The other silicon beam of each phase shifter is biased through the rest of the silicon structure of the fabric, which is taken as a ground. Phased array processors of 2×2 and 4×4 multiple-input-multiple-output (MIMO) ports are conveniently designed within reasonable footprints native to the current fabrication technologies. The response of the single-stage 4×4 broadband IPPAS element is determined, and its phase synthesis states required for single-throw, double-throw and broadcast routing operations are predicted. The transmission responses of the single-stage wavelength division multiplexing (WDM) processors of 2×2 and 4×4 MIMO ports are simulated. The wavelength steering capability of the transmission interferograms by applying progressive phase shifts through the array of NEMS-operated phase shift elements of the single-stage 4×4 WDM (de)multiplexer is demonstrated. The advantages of cascading broadband and WDM phased array sections are articulated through several study cases. Five different cascaded phased array architectures are trialed for the construction of non-blocking 4×4 IPPAS broadband switches that are essential elements in the construction of universal photonic processors. A cascaded 2×2 WDM (de)multiplexer that can set the bandwidth of the (de)multiplexed cyclic channels into a binary number of programmable values is demonstrated. The envelope and wavelength modulations of the transmission responses utilizing a cascaded forward structure of three 2×2 sections that can be utilized for the (de)multiplexing of different bandwidth channels are demonstrated providing individual wavelength steering capability of the narrowband and wideband channels and the individual wavelength steering capability of the slow envelope and wavelength modulating functions. Innovative universal 2×2 and 4×4 cascaded phased array processors of advanced high-order architectures that can function as both non-blocking broadband routers and tunable WDM (de)multiplexers with spectrum steering and bandwidth control of the (de)multiplexed demands are introduced. The multimode interference (MMI) coupler is utilized for the construction of several IIR feedback photonic processors. Tunable photonic feedback processors have the advantage of using less number of MMI couplers compared to their counterparts of FIR forward-path processors saving on the footprint and loss merits. A passive feedback 2×2 (de)multiplexer made of a 4×4 MMI coupler and two loopback paths is proposed. The inclusion of an imbalance in the lengths of the loopback paths of the same symmetrical feedback (de)multiplexer is demonstrated to achieve wavelength modulation of the (de)multiplexed transmission responses that are useful for the (de)multiplexing of different bandwidth channels. Several newly introduced IIR feedback architectures are demonstrated to function similarly as their counterparts of FIR forward-path processors as binary bandwidth variable (de)multiplexers, envelope and wavelength modulation (de)multiplexers, and universal feedback processors. The investigation provided in this thesis is also supported with dynamic zero-pole evolution analysis in the complex plane of analysis of the studied FIR and IIR photonic processors to enhance understanding the principle of operation. This research expands the prospective for constructing innovative silicon-on-insulator (SOI) based optical processors for applications in modern optical communication systems and programmable elastic optical networks (EONs).

Les réseaux optiques dynamiques et flexibles font partie des scénarios d'évolution des réseaux de transport optique. Ceux-ci formeront la base de la nouvelle génération des réseaux optiques de demain et permettront le déploiement efficace des services tel que le Cloud Computing. Cette évolution est destinée à apporter flexibilité et automatisation à la couche optique, mais s'accompagne d'une complexité supplémentaire, notamment au niveau de la gestion et de la commande de cette toute nouvelle génération de réseau. Jusqu'à récemment, les protocoles de routage et de signalisation normalisés ont pris en compte plusieurs paramètres physiques tels que l'information spectrale de la bande passante, le format de modulation, et la régénération optique. Cependant, d'autres paramètres sont encore nécessaires (par exemple, les puissances optiques des liens, le gain des amplificateurs) afin de faire fonctionner efficacement de grands réseaux. Dans ce contexte, il y a un besoin d'étudier les réseaux optiques existants ainsi que les différentes méthodes de prise en compte de la couche photonique dans le plan de contrôle. Le but est d'avoir un réseau optique automatique, flexible et programmable, mais surtout efficace de point de vue économique et opérationnel. L'utilisation de la technologie à grille flexible a un impact sur les réseaux optiques existants, où presque tous les équipements devront être remplacés, ce qui entraînera un coût additionnel pour les opérateurs. Dans ce travail, nous étudions les réseaux optiques actuels et évaluons l'impact de la flexibilité sur les infrastructures existantes. Ensuite, nous identifions plusieurs paramètres optiques à contrôler et proposons des extensions protocolaires afin d'intégrer ces paramètres dans un plan de contrôle GMPLS. De plus, nous développons les algorithmes de routage et de signalisation qui permettent la mise en œuvre d'un plan de contrôle efficace qui répond au besoin de la flexibilité. Enfin, l'ensemble de nos propositions et de nos solutions sont évaluées sur plusieurs topologies réseaux avec des modèles de trafic différents dans le but de valider leur pertinence. / Dynamic and flexible optical networks are among the evolution scenarios of the optical transport networks. These form the basis of the new generation of optical networks of tomorrow and enable the effective deployment of services such as cloud computing. This evolution is intended to provide flexibility and automation to the optical layer. However, it results in additional complexity, particularly in terms of the management and control of this new network generation. Until recently, the standardized routing and signaling protocols have been taking into account several optical parameters like the spectral bandwidth information, modulation format, and optical regeneration. However, other parameters (e.g., link optical powers, gain of optical amplifiers) are still required in order to efficiently operate large optical networks. In this context, there is a need to study the existing optical networks and the different integration methods of the photonic layer in a control plane. The goal is to get an automatic optical network that is flexible, programmable, and at the same time efficient from an economical and operational perspective. The use of flexible grid technology has an impact on existing optical networks, where almost all the equipment must be replaced, resulting in an additional cost to network operators. In this work, we study the current optical networks and evaluate the impact of flexibility on the existing infrastructures. Then, we identify several physical parameters to be controlled and propose protocol extensions in order to integrate these parameters in the GMPLS control plane. In addition, we develop the routing and signaling algorithms that allow the implementation of an efficient control plane that addresses the need for flexibility. Finally, the set of our proposals and solutions are evaluated on multiple network topologies with different traffic patterns in order to validate their relevance.