Autonomous and reliable operation of multilayer optical networks
This Ph.D. thesis focuses on the reliable autonomous operation of multilayer optical networks. The first objective focuses on the reliability of the optical network and proposes methods for health analysis related to Quality of Transmission (QoT) degradation. Such degradation is produced by soft-fai...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2022 |
| País: | España |
| Institución: | CBUC, CESCA |
| Repositorio: | TDR. Tesis Doctorales en Red |
| OAI Identifier: | oai:www.tdx.cat:10803/674551 |
| Acceso en línea: | http://hdl.handle.net/10803/674551 https://dx.doi.org/10.5821/dissertation-2117-369086 |
| Access Level: | acceso abierto |
| Palabra clave: | Àrees temàtiques de la UPC::Informàtica 004 621.3 |
| Sumario: | This Ph.D. thesis focuses on the reliable autonomous operation of multilayer optical networks. The first objective focuses on the reliability of the optical network and proposes methods for health analysis related to Quality of Transmission (QoT) degradation. Such degradation is produced by soft-failures in optical devices and fibers in core and metro segments of the operators’ transport networks. Here, we compare estimated and measured QoT in the optical transponder by using a QoT tool based on GNPy. We show that the changes in the values of input parameters of the QoT model representing optical devices can explain the deviations and degradation in performance of such devices. We use reverse engineering to estimate the value of those parameters that explain the observed QoT. We show by simulation a large anticipation in soft-failure detection, localization and identification of degradation before affecting the network. Finally, for validating our approach, we experimentally observe the high accuracy in the estimation of the modeling parameters. The second objective focuses on multilayer optical networks, where lightpaths are used to connect packet nodes thus creating virtual links (vLink). Specifically, we study how lightpaths can be managed to provide enough capacity to the packet layer without detrimental effects in their Quality of Service (QoS), like added delays or packet losses, and at the same time minimize energy consumption. Such management must be as autonomous as possible to minimize human intervention. We study the autonomous operation of optical connections based on digital subcarrier multiplexing (DSCM). We propose several solutions for the autonomous operation of DSCM systems. In particular, the combination of two modules running in the optical node and in the optical transponder activate and deactivate subcarriers to adapt the capacity of the optical connection to the upper layer packet traffic. The module running in the optical node is part of our Intent-based Networking (IBN) solution and implements prediction to anticipate traffic changes. Our comprehensive study demonstrates the feasibility of DSCM autonomous operation and shows large cost savings in terms of energy consumption. In addition, our study provides a guideline to help vendors and operators to adopt the proposed solutions. The final objective targets at automating packet layer connections (PkC). Automating the capacity required by PkCs can bring further cost reduction to network operators, as it can limit the resources used at the optical layer. However, such automation requires careful design to avoid any QoS degradation, which would impact Service Level Agreement (SLA) in the case that the packet flow is related to some customer connection. We study autonomous packet flow capacity management. We apply RL techniques and propose a management lifecycle consisting of three different phases: 1) a self-tuned threshold-based approach for setting up the connection until enough data is collected, which enables understanding the traffic characteristics; 2) RL operation based on models pre-trained with generic traffic profiles; and 3) RL operation based on models trained with the observed traffic. We show that RL algorithms provide poor performance until they learn optimal policies, as well as when the traffic characteristics change over time. The proposed lifecycle provides remarkable performance from the starting of the connection and it shows the robustness while facing changes in traffic. The contribution is twofold: 1) and on the one hand, we propose a solution based on RL, which shows superior performance with respect to the solution based on prediction; and 2) because vLinks support packet connections, coordination between the intents of both layers is proposed. In this case, the actions taken by the individual PkCs are used by the vLink intent. The results show noticeable performance compared to independent vLink operation. |
|---|