Resumen de Cognition procedures for optical network design and optimization
Telecom carriers have to adapt their networks to accommodate a growing volume of users, services and traffic. Thus, they have to search a continuous maximization of efficiency and reduction in costs. This thesis identifies an opportunity to accomplish this aim by reducing operation margins applied in the optical link power budgets, in optical transport networks. From an operational perspective, margin reduction will lead to a fall of the required investments on transceivers in the whole transport network. Based on how human learn, a cognitive approach is introduced and evaluated to reduce the System Margin. This operation margin takes into account, among other constraints, the long-term ageing process of the network infrastructure. Telecom operators normally apply a conservative and fixed value established during the design and commissioning phases. The cognitive approach proposes a flexible and variable value, adapted to the network conditions. It is based on the case-based reasoning machine learning technique, which has been further developped. Novel learning schemes are presented and evaluated. The cognition solution proposes a new lower launched power guaranteeing the quality of service of the new incoming lightpath. It will lead to provide transmission power savings with appropiate success rates when applying the cognitive approach. To this end, it relies on transmission values applied in past and successful similar network situations. They are stored in a knowledge base or memory of the system. Moreover, regarding the knowledge base, a static and a dynamic approaches have been developped and presented. In the last case, five new dynamic learning algorithms are presented and evaluated. In the static context, savings in transmission power up to 48% are achieved and the resulting System Margin reduction. Furthermore, the dynamic renewal of the knowledge base improves mean savings in launched power up to 7% or 18% with respect to the static approach, depending on the path. Thus, the cognitive approach appears as useful to be applied in commercial optical transport networks with the aim of reducing the operational System Margin.
