As redes atuais buscam a integração dos serviços e tendem ao IP/MPLS sobre DWDM. No futuro, espera-se que as redes sejam do tipo GMPLS, apresentando melhoria na flexibilidade, capacidade de comutação no meio óptico e também um plano de controle único. Tais redes poderão prover a integração de diferentes camadas e tecnologias, além de reduzirem os custos de operação e de provisionamento. Dentre os muitos aspectos desta nova tecnologia, o trabalho proposto concentra-se no estudo do roteamento óptico em redes GMPLS, verificando a relação existente entre algoritmos RWA e o dimensionamento das mesmas. Pesquisas mostram que, ao tentar estabelecer uma rota numa rede totalmente óptica e com tráfego dinâmico, o bloqueio de conexões e o número de falhas podem limitar o seu desempenho. Realizando-se simulações com algoritmos RWA, é possível determinar o número mínimo de comprimentos de onda e avaliar a melhor topologia de rede para uma determinada probabilidade de bloqueio de conexões e de falhas. Este trabalho mostra como simulações com algoritmos RWA auxiliam no dimensionamento de redes GMPLS permitindo determinar a influência destes algoritmos em seu desempenho. Como principais resultados estão o dimensionamento de recursos, a determinação da carga de tráfego de trabalho e da taxa de falha permitida e a seleção do melhor tipo de algoritmo RWA para a rede de pesquisa Kyatera em duas possíveis fases de sua implementação. / Today's networks seek integration of services and tend towards lean IP/MPLS over DWDM. In the future, it is expected that the networks will tend towards GMPLS with enhanced flexibility and switching capability in the optical layer and a unified control plane. Such networks will provide integration between different network layers and technologies, besides decrease operating and provisioning costs. Among many aspects of this technology, this study concentrates on GMPLS network optical routing, verifying the relationship between RWA algorithms and network dimensioning. Research shows that when attempting to establish a route in an alloptical network with dynamic traffic, connection blocking and failures can limit performance. Through of RWA algorithms simulation, it is possible to determine the minimum wavelength number and check the best network topology for a given probability of connection blocking and failure. This work shows how RWA simulations can assist in the dimensioning of GMPLS networks and in determining the influence of RWA algorithm on their performance. The main results are resource dimensioning, the determination of working traffic load and the allowed failure rate and the selection of the best RWA algorithm for Kyatera research network in two possible phases of its implementation.

As redes atuais buscam a integração dos serviços e tendem ao IP/MPLS sobre DWDM. No futuro, espera-se que as redes sejam do tipo GMPLS, apresentando melhoria na flexibilidade, capacidade de comutação no meio óptico e também um plano de controle único. Tais redes poderão prover a integração de diferentes camadas e tecnologias, além de reduzirem os custos de operação e de provisionamento. Dentre os muitos aspectos desta nova tecnologia, o trabalho proposto concentra-se no estudo do roteamento óptico em redes GMPLS, verificando a relação existente entre algoritmos RWA e o dimensionamento das mesmas. Pesquisas mostram que, ao tentar estabelecer uma rota numa rede totalmente óptica e com tráfego dinâmico, o bloqueio de conexões e o número de falhas podem limitar o seu desempenho. Realizando-se simulações com algoritmos RWA, é possível determinar o número mínimo de comprimentos de onda e avaliar a melhor topologia de rede para uma determinada probabilidade de bloqueio de conexões e de falhas. Este trabalho mostra como simulações com algoritmos RWA auxiliam no dimensionamento de redes GMPLS permitindo determinar a influência destes algoritmos em seu desempenho. Como principais resultados estão o dimensionamento de recursos, a determinação da carga de tráfego de trabalho e da taxa de falha permitida e a seleção do melhor tipo de algoritmo RWA para a rede de pesquisa Kyatera em duas possíveis fases de sua implementação. / Today's networks seek integration of services and tend towards lean IP/MPLS over DWDM. In the future, it is expected that the networks will tend towards GMPLS with enhanced flexibility and switching capability in the optical layer and a unified control plane. Such networks will provide integration between different network layers and technologies, besides decrease operating and provisioning costs. Among many aspects of this technology, this study concentrates on GMPLS network optical routing, verifying the relationship between RWA algorithms and network dimensioning. Research shows that when attempting to establish a route in an alloptical network with dynamic traffic, connection blocking and failures can limit performance. Through of RWA algorithms simulation, it is possible to determine the minimum wavelength number and check the best network topology for a given probability of connection blocking and failure. This work shows how RWA simulations can assist in the dimensioning of GMPLS networks and in determining the influence of RWA algorithm on their performance. The main results are resource dimensioning, the determination of working traffic load and the allowed failure rate and the selection of the best RWA algorithm for Kyatera research network in two possible phases of its implementation.

Η αλματώδης ανάπτυξη που παρατηρείται τα τελευταία χρόνια στον τομέα των οπτικών δικτύων τηλεπικοινωνιών, έχει ως συνέπεια την αύξηση της κατανάλωσης ενέργειας. Σήμερα υπολογίζεται ότι η ενέργεια που οφείλεται σε IT τεχνολογίες αποτελεί το 1%-2% της παγκόσμιας κατανάλωσης. Ωστόσο, τα ποσοστά αυτά αναμένεται να αυξηθούν δραματικά μιας και όλο και περισσότερες εφαρμογές, που απαιτούν μεγάλο εύρος ζώνης (π.χ. video streaming, HDTV), αναπτύσσονται ταχύτατα. Μέχρι πρόσφατα ο κύριος περιορισμός στην ανάπτυξη των δικτύων ήταν το κόστος εξοπλισμού και μετάδοσης. Σήμερα το μεγαλύτερο πρόβλημα είναι η όλο και αυξανόμενη κατανάλωση ενέργειας. Έτσι λοιπόν πιο energy-aware προσεγγίσεις για το σχεδιασμό, την υλοποίηση και τη λειτουργία των οπτικών δικτύων κρίνονται πλέον απαραίτητες. Για την αντιμετώπιση του προβλήματος της κατανάλωσης ενέργειας σε οπτικά δίκτυα υπάρχουν δυο προσεγγίσεις: η ανάπτυξη ενεργο-αποδοτικών συσκευών δικτύου ή η ανάπτυξη αλγορίθμων δρομολόγησης και ανάθεσης μήκους κύματος (RWA) που να λαμβάνουν υπόψη τους και την ενέργεια που καταναλώνει το δίκτυο. Στην παρούσα διπλωματική εργασία επιλέχθηκε η δεύτερη προσέγγιση όπου και αναπτύχθηκαν Energy Aware RWA αλγόριθμοι για τον σχεδιασμό WDM οπτικών δικτύων με στόχο την ελαχιστοποίηση της ενέργειας που καταναλώνεται. Αυτό επιτυγχάνεται μειώνοντας τον αριθμό των συσκευών του δικτύου, όπως αναγεννητές, ενισχυτές, add/drop τερματικά κ.τ.λ. που είναι γνωστό ότι καταναλώνουν αρκετή ενέργεια κατά τη λειτουργία τους. Αρχικά αναπτύχθηκε ένας αλγόριθμος σε Ακέραιο Γραμμικό Ακέραιο Προγραμματισμό (Integer Linear Programming - ILP) που ελαχιστοποιεί από κοινού όλες εκείνες τις συσκευές του δικτύου που καταναλώνουν ενέργεια. Έπειτα το πρόβλημα χωρίζεται σε δυο υπο-προβλήματα: την κατανομή αναγεννητών σε κόμβους του δικτύου όπου η αναγέννηση είναι απαραίτητη και στο Energy Aware RWA. Σε αυτή τη περίπτωση εκτός από ILP προγραμματισμό εφαρμόζεται επίσης και χαλάρωση των περιορισμών ακεραιότητας (LP προγραμματισμός) ώστε να μπορούν να προκύψουν λύσεις και για μεγαλύτερες τοπολογίες. Για τον υπολογισμό της απόδοσης των προτεινόμενων αλγορίθμων διεξάγονται μια σειρά από πειράματα εξομοιώσεων σε Matlab, με τη βοήθεια του LINDO API για την επίλυση των (I)LP προβλημάτων σε ένα μικρό δίκτυο αλλά και στο ευρωπαϊκό δίκτυο Geant-2. Oι προτεινόμενοι αλγόριθμοι συγκρίνονται με τυπικούς RWA αλγορίθμους που δεν λαμβάνουν υπόψη την ενέργεια ή ελαχιστοποιούν κάποιες από τις συνιστώσες του δικτύου. Οι υπολογισμοί δείχνουν ότι η κατανάλωση της ενέργειας του δικτύου μπορεί να μειωθεί μέχρι και 45% σε δίκτυο με συνδέσμους μονής ίνας και μέχρι και 65% σε συνδέσμους διπλής ίνας. / The continuing deployment and upgrade of optical telecommunication networks drive up power and energy consumption, in a way that makes operators worry that future energy consumption levels may pose constraints on communications growth that are more significant than those posed by bandwidth considerations. As community concerns about global energy consumption grow, the power consumption of the optical networks is becoming an issue of increasing importance. Even though energy consumption of IT-related infrastructure currently represents between 1% and 2% of global energy consumption, such values are destined to rise, as more bandwidth-hungry applications (such as video streaming, and HDTV in particular) continue to be developed. Thus, it seems that an energy-aware approach is increasingly needed during the design, implementation and operation of optical networks. Two different approaches can be explored to reduce energy consumption in optical networks: the improvement of the energy efficiency of the equipment and the energy awareness of the Routing and Wavelengths Assignment (RWA) algorithms. In this work the energy minimization problem in optical networks is considered from an algorithmic perspective. The objective of the proposed algorithms is to plan optical WDM networks so as to minimize the energy consumed, by minimizing the number of the most energy-consuming components. Such components can be amplifiers, regenerators, add/drop terminals, optical fibers, etc. Firstly an Integer Linear Programming (ILP) formulation is given that aims at jointly minimizing the energy consuming modules present in an optical network, such as the regenerators, amplifiers, wavelength selective switches, etc. Then the problem for a general (translucent network) is decomposed into a regeneration placement problem, and an energy-aware RWA for transparent networks problem, where again ILP formulations are given to minimize energy consumption. The ILP of the decomposed problem is solved as an LP problem, by relaxing the integer constraints, in order to obtain solutions for larger network topologies. In order to evaluate the performance of the proposed Energy Aware (EA)-RWA algorithms a number of simulation experiments is carried out. All the RWA algorithms are implemented in Matlab and LINDO API is used to solve the corresponding (I)LP problems. The proposed algorithms are compared with RWA algorithms that do not consider energy minimization techniques or RWA algorithms that partially take into account energy consuming components. The simulation results showed that the energy consumption in the optical layer can be reduced by up to 45% in single fiber networks and by up to 65% in multi-fiber networks.

Made available in DSpace on 2015-05-14T12:36:49Z (GMT). No. of bitstreams: 1 arquivototal.pdf: 2456764 bytes, checksum: 4ac348983da5718e79b705aa01b9aa0a (MD5) Previous issue date: 2014-08-15 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / This dissertation proposes two new algorithms for routing and wavelength allocation EA-RWA (Energy Aware Routing and Wavelength Assigment) and IEA-RWA (Impairment and Energy Aware Routing and Wavelength Assigment). The algorithms have the characteristic, improve energy efficiency in dynamic transparent optical networks, the first being simpler, because it does not care about the imperfections of the physical layer. The EA-RWA is compared with a traditional RWA / classic that does not take into account at the time of admission of the physical layer connections restrictions nor energy consumption algorithm. Have the IEA-RWA is compared with a RWA algorithm that takes into account the constraints (IA-RWA) physical layer. Numerical simulations show that the IEA-RWA performs better in many aspects / Esta dissertação propõe dois novos algoritmos de roteamento e alocação de comprimento de onda, a saber, o EA-RWA (Energy Aware Routing and Wavelength Assigment) e o IEA-RWA (Impairment and Energy Aware Routing and Wavelength Assigment). Os algoritmos terão como característica, melhorar a eficiência energética em redes ópticas transparentes dinâmicas, sendo o primeiro mais simples, pois não se preocupa com as imperfeições da camada física. O EA-RWA é comparado com um algoritmo RWA tradicional/clássico que não leva em conta, no momento de admissão de conexões, restrições da camada física e nem consumo de energia. Já o IEA-RWA é comparado com um algoritmo RWA que leva em conta restrições da camada física (IA-RWA). As simulações numéricas mostram que o IEA-RWA apresenta melhor desempenho em vários aspectos