Cross-layer optimization in optical networks
ColaboratorSpadaro, Salvatore; Careglio, Davide; Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors
Document typeDoctoral thesis
PublisherUniversitat Politècnica de Catalunya
Rights accessOpen Access
Network planning and operation in core optical networks require processes that lead to a cost-effective network able to effectively accommodate the given traffic demands. Crosslayer techniques that exploit information coming from the transport plane to serve either the network planning or the operation phases have been proposed to achieve optimal resource utilization and network performance. Dynamic impairment-aware networking refers to a solution that utilizes the dynamicity as well as the valuable physical-layer information of a reconfigurable WDM network to provide a smooth transition from the quasi-static networking of today to an intelligent reconfigurable and physical impairment-aware architecture. The concept of physical layer awareness allows intelligent techniques to offer optimal planning, dynamic configuration and management of the network while ensuring strong quality of transmission (QoT). A physical-layer performance estimating tool called Q-tool, has been designed and developed to deliver fast and accurate QoT assessments in dynamic single line-rate WDM networks employing 10 Gb/s OOK systems. The Q-tool was developed so as to feed with QoT evaluations the various cross-layer modules including offline RWA, monitor placement, regenerator placement, online RWA and failure handling. The significance of the physical-layer awareness and the role of Q-tool in the online routing strategy have been experimentally investigated using an impairment-enabled control-plane testbed. A cross-layer module that utilizes physical layer information to optimize the planning phase of single line-rate networks has been also developed; a monitor placement scheme was designed that takes into account partial monitoring information coming from the physical layer to decide the optimum number and locations of the optical monitoring devices. A techno-economic analysis was also conducted to explore the cost implications stemming from the resource optimization of a dynamic single line-rate network. Impairment aware and impairment-unaware algorithms along with the developed QoT tool were used to compare the planning solutions in terms of CapEx and OpEx. The challenges arising in the core networks of the next generation have also been addressed. Flexible optical networking has been introduced by the research community as a way to offer efficient utilization of the available optical resources while offering ultra-high speed rates. As opposed to the rate-specific and fixed-grid solution of a mixed line-rate (MLR) network, flexible networks are bandwidth agnostic and have the ability to adapt on- demand the delivered bit-rate. Yet, the increased level of flexibility imposes complex requirements; during network operation traffic is changing with time, leaving windows of spectrum of variable size unused. Moreover, interference from the neighbouring channels has to be taken into account and sufficient guard-bands ought to be considered. In such context the physical layer requirements for a spectrum-flexible optical superchannel were experimentally investigated by implementing a set of networking scenarios. The importance of the result is that it can act as input for cognitive algorithms that will enable a novel networking paradigm. Finally next generation core networks were evaluated from a cost, spectral and energy perspective so as to give a comprehensive view of the potential of the proposed technologies. The resource optimization achieved by flexible networks has been compared to single and mixed line-rate networks under the prism of cost and energy efficiency. It was shown that the capability of flexible networks to allocate efficiently the available spectrum, counterbalances the additional capital expenditures that are required to migrate to a multi-carrier system. In addition flexible networks achieve low energy per bit as they use just the amount of network resources needed for the given input traffic.
- Tesis - TDX-UPC