Le caching de réseau peut aider àgérer l'explosion du trafic sur Internet et àsatisfaire la Qualité d'Expérience (QoE)croissante demandée par les usagers.Néanmoins, les techniques proposées jusqu'àprésent par la littérature scientifique n'arriventpas à exploiter tous les avantages potentiels. Lestravaux de recherche précédents cherchent àoptimiser le hit ratio ou d'autres métriques deréseau, tandis que les opérateurs de réseau(ISPs) sont plus intéressés à des métriques plusconcrètes, par exemple le coût et la qualitéd'expérience (QoE). Pour cela, nous visonsdirectement l'optimisation des métriquesconcrètes et montrons que, ce faisant, on obtientdes meilleures performances.Plus en détail, d'abord nous proposons desnouvelles techniques de caching pour réduire lecoût pour les ISPs en préférant stocker lesobjets qui sont les plus chères à repérer.Nous montrons qu'un compromis existe entre lamaximisation classique du hit ratio et laréduction du coût.Ensuite, nous étudions la distribution vidéo,comme elle est la plus sensible à la QoE etconstitue la plus part du trafic Internet. Lestechniques de caching classiques ignorent sescaractéristiques particulières, par exemple le faitqu'une vidéo est représentée par différentesreprésentations, encodées en différents bit-rateset résolutions. Nous introduisons des techniquesqui prennent en compte cela.Enfin, nous remarquons que les techniquescourantes assument la connaissance parfaite desobjets qui traversent le réseau. Toutefois, laplupart du trafic est chiffrée et du coup toutetechnique de caching ne peut pas fonctionner.Nous proposons un mécanisme qui permet auxISPs de faire du caching, bien qu’ils ne puissentobserver les objets envoyés. / Network caching can help copewith today Internet traffic explosion and sustainthe demand for an increasing user Quality ofExperience. Nonetheless, the techniquesproposed in the literature do not exploit all thepotential benefits. Indeed, they usually aim tooptimize hit ratio or other network-centricmetrics, e.g. path length, latency, etc., whilenetwork operators are more focused on moremore practical metrics, like cost and quality ofexperience. We devise caching techniques thatdirectly target the latter objectives and showthat this allows to gain better performance.More specifically, we first propose novelstrategies that reduce the Internet ServiceProvider (ISP) operational cost, bypreferentially caching the objects whose cost ofretrieval is the largest.We then focus on video delivery, since it is themost sensitive to QoE and represents most ofthe Internet traffic. Classic techniques ignorethat each video is represented by differentrepresentations, encoded at different bit-ratesand resolutions. We devise techniques that takethis into account.Finally, we point out that the techniquespresented in the literature assume the perfectknowledge of the objects that are crossing thenetwork. Nonetheless, most of the traffic todayis encrypted and thus caching techniques areinapplicable. To overcome this limit, Wepropose a mechanism which allows the ISPs tocache, even without knowing the objects being

The objective of the research is to study optimal routing policy (ORP) problems and to develop an optimal adaptive routing algorithm practical for large-scale Stochastic Time-Dependent (STD) real-life networks, where a traveler could revise the route choice based upon en route information. The routing problems studied can be viewed as counterparts of shortest path problems in deterministic networks. A routing policy is defined as a decision rule that specifies what node to take next at each decision node based on realized link travel times and the current time. The existing routing policy algorithm is for explorative purpose and can only be applied to hypothetical simplified network. In this research, important changes have been made to make it practical in a large-scale real-life network. Important changes in the new algorithm include piece-wise linear travel time representation, turn-based, label-correcting, criterion of stochastic links, and dynamic blocked links. Complete dependency perfect online information (CDPI) variant is then studied in a real-life network (Pioneer Valley, Massachusetts). Link travel times are modeled as random variables with time-dependent distributions which are obtained by running Dynamic Traffic Assignment (DTA) using data provided by Pioneer Valley Planning Commission (PVPC). A comprehensive explanation of the changes by comparing the two algorithms and an in-depth discussion of the parameters that affects the runtime of the new algorithm is given. Computational tests on the runtime changing with different parameters are then carried out and the summary of its effectiveness are presented. To further and fully understand the applicability and efficiency, this algorithm is then tested in another large-scale network, Stockholm in Sweden, and in small random networks. This research is also a good starting point to investigate strategic route choice models and strategic route choice behavior in a real-life network. The major tasks are to acquire data, generate time-adaptive routing policies, and estimate the runtime of the algorithm by changing the parameters in two large-scale real-life networks, and to test the algorithm in small random networks. The research contributes to the knowledge base of ORP problems in stochastic time-dependent (STD) networks by developing an algorithm practical for large-scale networks that considers complete time-wise and link-wise stochastic dependency.

Evolutionary algorithms have achieved great performance among a wide range of optimization problems. In this degree project, the network optimization problem has been reformulated and solved in an evolved way. A feasible evolutionary framework has been designed and implemented to train neural networks in supervised learning scenarios. Under the structure of evolutionary algorithms, a well-defined fitness function is applied to evaluate network parameters, and a carefully derived form of approximate gradients is used for updating parameters. Performance of the framework has been tested by training two different types of networks, linear affine networks and convolutional networks, for a flash call verification task.Under this application scenario, whether a flash call verification will be successful or not will be predicted by a network, which is inherently a binary classification problem. Furthermore, its performance has also been compared with traditional backpropagation optimizers from two aspects: accuracy and time consuming. The results show that this framework is able to push a network training process to converge into a certain level. During the training process, despite of noises and fluctuations, both accuracies and losses converge roughly under the same pattern as in backpropagation. Besides, the evolutionary algorithm seems to have higher updating efficiency per epoch at the first training stage before converging. While with respect to fine tuning, it doesn’t work as good as backpropagation in the final convergence period. / Evolutionära algoritmer uppnår bra prestanda för ett stort antal olika typer av optimeringsproblem. I detta examensprojekt har ett nätverksoptimeringsproblem lösts genom omformulering och vidareutveckling av angreppssättet. Ett förslag till ramverk har utformats och implementerats för att träna neuronnätverk i övervakade inlärningsscenarier. För evolutionära algoritmer används en väldefinierad träningsfunktion för att utvärdera nätverksparametrar, och en noggrant härledd form av approximerade gradienter används för att uppdatera parametrarna. Ramverkets prestanda har testats genom att träna två olika typer av linjära affina respektive konvolutionära neuronnätverk, för optimering av telefonnummerverifiering. I detta applikationsscenario förutses om en telefonnummerverifiering kommer att lyckas eller inte med hjälp av ett neuronnätverk som i sig är ett binärt klassificeringsproblem. Dessutom har dess prestanda också jämförts med traditionella backpropagationsoptimerare från två aspekter: noggrannhet och hastighet. Resultaten visar att detta ramverk kan driva en nätverksträningsprocess för att konvergera till en viss nivå. Trots brus och fluktuationer konvergerar både noggrannhet och förlust till ungefär under samma mönster som i backpropagation. Dessutom verkar den evolutionära algoritmen ha högre uppdateringseffektivitet per tidsenhet i det första träningsskedet innan den konvergerar. När det gäller finjustering fungerar det inte lika bra som backpropagation under den sista konvergensperioden.

No description available.

Wireless operators adopting third-generation (3G) technologies and those migrating from second-generation (2G) to 3G face a number of challenges related to traffic modeling, demand characterization, and performance analysis, which are key elements in the processes of designing, dimensioning and optimizing their network infrastructure. Traditional traffic modeling assumptions used for circuit-switched voice traffic no longer hold true with the convergence of voice and data over packet-switched infrastructures. Self-similar models need to be explored to appropriately account for the burstiness that packet traffic is expected to exhibit in all time scales. The task of demand characterization must include an accurate description of the multiple user profiles and service classes the network is expected to support, with their distinct geographical distributions, as well as forecasts of how the market should evolve over near and medium terms. The appropriate assessment of the quality of service becomes a more complex issue as new metrics and more intricate dependencies have to be considered when providing a varying range of services and applications that include voice, real-time, and non-real time data. All those points have to be considered by the operator to obtain a proper dimensioning, resource allocation, and rollout plan for system deployment. Additionally, any practical optimization strategy has to rely on accurate estimates of expected demand and growth in demand. In this research, we propose a practical framework to characterize the traffic offered to multimedia wireless systems that allows proper dimensioning and optimization of the system for a particular demand scenario. The framework proposed includes a methodology to quantitatively and qualitatively describe the traffic offered to multimedia wireless systems, solutions to model that traffic as practical inputs for simulation analysis, and investigation of demand-sensitive techniques for system dimensioning and performance optimization. We consider both theoretical and practical aspects related to the dimensioning of hybrid traffic (voice and data) for mobile wireless networks. We start by discussing wireless systems and traffic theory, with characterization of the main metrics and models that describe the users’ voice and data demand, presenting a review of the most recent developments in the area. The concept of service class is used to specify parameters that depend on the application type, performance requirements and traffic characteristics for a given service. Then we present the concept of “user profile,“ which ties together a given combination of service class, propagation environment and terminal type. Next, we propose a practical approach to explore the dynamics of user geographical distribution in creating multi-service, multi-class traffic layers that serve as input for network traffic simulation algorithms. The concept of quality-of-service (QoS) is also discussed, focusing on the physical layer for 3G systems. We explore system simulation as a way to dimension a system given its traffic demand characterization. In that context, we propose techniques to translate geographical distributions of user profiles into the actual number of active users of each layer, which is the key parameter to be used as input in simulations. System level simulations are executed for UMTS systems, with the purpose of validating the methodology proposed here. We complete the proposed framework by applying all elements together in the process of dimensioning and optimization of 3G wireless networks using the demand characterization for the system as input. We investigate the effects of modifying some elements in the system configuration such as network topology, radio-frequency (RF) configuration, and radio resource management (RRM) parameters, using strategies that are sensitive to traffic geographical distribution. Case study simulations are performed for Universal Mobile Telecommunications System (UMTS) networks, and multiple system variables (such as antenna tilts, pilot powers, and RRM parameters) are optimized using traffic sensitive strategies, which result in significant improvements in the overall system capacity and performance. Results obtained in the case studies, allied to a generic discussion of the trade-offs involved in the proposed framework, demonstrate the close dependence between the processes of system dimensioning and optimization with the accurate modeling of traffic demand offered to the system. / Ph. D.

The distributed and privacy sensitive nature of cellular networks make them strong candidates for optimization using Federated Learning, but this exposes them to a problem inherent to the learning paradigm: performance inequality due to heterogeneous client data distributions. The prevailing approach of enforcing uniform client performance ignores client-specific performance limitations due to different levels of irreducible uncertainty present in their data, resulting in deteriorated network performance. To address this issue, this thesis introduces two novel federated algorithms designed to enhance learning efficiency and ensure fairness in the presence of heteroscedastic noise, reflecting the distributive justice principles of utilitarianism and equality. Under these circumstances, the proposed algorithms are shown to significantly improve overall performance and performance fairness. The deployment of these algorithms promises a dual benefit: enhancement in network performance and a fairer distribution of service quality for end users.

Detta examensarbete handlar om hur datatrafik över en satellitlänk kan optimeras för att minska laddningstider och överförd datamängd. Syftet med studien är att undersöka i vilken omfattning datatrafik mellan fartyg och land via satellitlänk kan styras så att trafiken blir effektivare. Genom att använda DNS-mellanlagring, mellanlagring av webbsidor samt annonsblockering med pfSense som plattform har examensarbetet utfört experiment emot olika hemsidor och mätt laddningstid samt överförd datamängd. Resultatet visade att det fanns stora möjligheter att optimera nätverkstrafiken och de uppmätta resultaten visade på en minskning av datamängden med 94% och laddningstiderna med 67%. / The thesis describes how network traffic transmitted via a satellite link can be optimized in order to reduce loading times and transmitted data. The purpose with this study has been to determine what methods are available to control and reduce the amount of data transmitted through a network and how this data is affected. By applying the practice of DNS caching, web caching and ad blocking with the use of pfSense as a platform the study has performed experiments targeting different web sites and measured the loading times and amount of transmitted data. The results showed good possibilities to optimize the network traffic and the measured values indicated a reduction of the network traffic of up to 94% and loading times with 67%.

Selon l'Union Internationale des Télécommunications (UIT), la progression importante du nombre de téléphones mobiles à travers le monde a dépassé toutes les prévisions avec un nombre d'utilisateurs estimé à 6 Mds en 2011 dont plus de 75% dans les pays développés. Cette progression importante produit une pression forte sur les opérateurs de téléphonie mobile concernant les ressources radio et leur impact sur la qualité et le degré de service (GoS) dans le réseau. Avec des demandes différenciées de services émanant de différentes classes d'utilisateurs, la capacité d'identifier les types d'utilisateurs dans le réseau devient donc vitale pour l'optimisation de l'infrastructure et des ressources. Dans la présente thèse, une nouvelle approche de classification des utilisateurs d'un réseau cellulaire mobile est proposée, en exploitant les données du trafic réseau fournies par deux opérateurs de téléphonie mobile en Afrique du Sud. Dans une première étape, celles-ci sont décomposées en utilisant deux méthodes multi-échelles ; l'approche de décomposition en mode empirique (Empirical Mode Decomposition approach - EMD) et l'approche en Ondelettes Discrètes (Discrete Wavelet Packet Transform approach - DWPT). Les résultats sont ensuite comparés avec l'approche dite de Difference Histogram qui considère le nombre de segments de données croissants dans les séries temporelles. L'approche floue de classification FCM (Fuzzy C-means) est utilisée par la suite pour déterminer les clusters, ou les différentes classes présentes dans les données, obtenus par analyse multi-échelles et par différence d'histogrammes. Les résultats obtenus montrent, pour la méthode proposée, une séparation claire entre les différentes classes de trafic par rapport aux autres méthodes. La deuxième partie de la thèse concerne la proposition d'une approche d'optimisation des ressources réseau, qui prend en compte la variation de la demande en termes de trafic basée sur les classes d'abonnés précédemment identifiés dans la première partie. Une nouvelle approche hybride en deux niveaux pour l'allocation des canaux est proposée. Le premier niveau considère un seuil fixe de canaux alloués à chaque cellule en prenant en considération la classe d'abonnés identifiée par une stratégie statique d'allocation de ressources tandis que le deuxième niveau considère une stratégie dynamique d'allocation de ressources. Le problème d'allocation de ressources est formulé comme un problème de programmation linéaire mixte (Mixed-Integer Linear programming - MILP). Ainsi, une approche d'allocation par période est proposée dans laquelle un groupe de canaux est alloué de façon dynamique pour répondre à la variation de la demande dans le réseau. Pour résoudre le problème précédent, nous avons utilisé l'outil CPLEX. Les résultats obtenus montrent qu'une solution optimale peux être atteinte par l'approche proposée (MILP) / The growth in the number of cellular mobile subscribers worldwide has far outpaced expected rates of growth with worldwide mobile subscriptions reaching 6 Billion subscribers in 2011 according to the International Telecommunication Union (ITU). More than 75% of this figure is in developing countries. With this rate of growth, greater pressure is placed on radio resources in mobile networks which impacts on the quality and grade of service (GOS) in the network. With varying demands that are generated from different subscriber classes in a network, the ability to distinguish between subscriber types in a network is vital to optimise infrastructure and resources in a mobile network. In this study, a new approach for subscriber classification in mobile cellular networks is proposed. In the proposed approach, traffic data extracted from two network providers in South Africa is considered. The traffic data is first decomposed using traditional feature extraction approaches such as the Empirical Mode Decomposition (EMD) and the Discrete Wavelet Packet Transform (DWPT) approach. The results are then compared with the Difference Histogram approach which considers the number of segments of increase in the time series. Based on the features extracted, classification is then achieved by making use of a Fuzzy C-means algorithm. It is shown from the results obtained that a clear separation between subscriber classes based on inputted traffic signals is possible through the proposed approach. Further, based on the subscriber classes extracted, a novel two-level hybrid channel allocation approach is proposed that makes use of a Mixed Integer Linear Programming (MILP) model to consider the optimisation of radio resources in a mobile network. In the proposed model, two levels of channel allocation are considered: the first considers defining a fixed threshold of channels allocated to each cell in the network. The second level considers a dynamic channel allocation model to account for the variations in traffic experienced in each traffic class identified. Using the optimisation solver, CPLEX, it is shown that an optimal solution can be achieved with the proposed two-level hybrid allocation model

La crise énergétique marquée par une flambée des prix du pétrole et les impératifs d’un développement durable font des énergies renouvelables une alternative qui suscitent aujourd'hui l’intérêt de plusieurs équipes de recherches. Le Sénégal, pays subsaharien, non producteur de pétrole n’est pas épargné par cette crise. Celle-ci se traduit par, une faible couverture du pays (en particulier les zones rurales) par le réseau électrique national. Pour faire face à cela, la nouvelle orientation en matière de politique énergétique met l’accent sur l’utilisation des énergies renouvelables notamment, le solaire et l’éolienne. L'objectif principal de cette thèse est de dégager une méthodologie de conception d’un site isolé de production d’énergie électrique à partir des énergies alternatives. Il s'agit à partir des caractéristiques d’un site (ressources d’énergie disponibles, besoins énergétiques) de :- mettre en place une démarche qui permette le choix d’une architecture du réseau et de dimensionner de façon optimale l’ensemble des constituants du réseau (machines, sources et dispositifs de stockage) en tenant compte de toutes les contraintes- concevoir un dispositif de commande des composantes et de gestion des flux d’énergie Cette thèse comprend trois chapitres :- Chapitre 1 : pose la problématique de l’énergie au Sénégal. L’organisation institutionnelle du sous-secteur de l’électricité, les réalisations et les projets en cours sont présentés. Aussi, une étude sur les architectures de micro-réseaux est faite. - Chapitre 2 : traite des outils et logiciels. Une étude comparative des principaux logiciels d’analyse, de conception et simulation des micro-réseaux est réalisée. L’ensemble des composants de notre système ont été modélisées. Les méthodes d’optimisation et des outils de représentation graphique (Bond Graph, GIC et REM) des systèmes ont été présentés. Une enquête menée et une recherche bibliographique nous ont permis d’évaluer le potentiel énergétique du site et les besoins des populations.- Chapitre 3 : il s’agit de l’application de notre outil sur un site isolé identifié au Sénégal (MBoro/Mer). L’optimisation à travers la fonction objectif coût annualisée du système (ACS) nous a permis de dimensionner de façon optimale notre système. Aussi la commande du système avec la Représentation Energétique Macroscopique (REM) a été conçue.Pour la suite du travail, il faudrait envisager une prise en charge des problèmes de disponibilité du système à travers une surveillance et une supervision du dispositif. Le volet socio-économique aussi est à intégrer dans le futur afin de satisfaire l’évolution des besoins et des habitudes des populations. / The energy crisis characterized by the oil products price rising and the imperatives of sustainable development do that renewable energies are an alternative today witch attract the interest of several research teams. Senegal, sub-Saharan country, not oil producer is not spared by this crisis. The consequence of this is a low coverage of the country (especially rural areas) by the national grid. For solving this, the new orientation of the energy policy focuses on the use of renewable energy particularly solar and wind.The main objective of this thesis is to identify a methodology of design of an isolated site of electrical energy production from alternative energies It is consists on, from site characteristics (energy resources, energy requirements):- to develop an approach that allows the choice of network architecture and sizing optimally all components of the network (machines, sources and storage devices) taking into account all the constraints- to design a device for controlling components and managing the energy flowsThis thesis contains three chapters:- Chapter 1 - poses the problem of energy in Senegal. The institutional organization of the electricity sub-sector and the ongoing projects are presented. Also, a study of micro- architectures networks is made.- Chapter 2 deals with tools and software. A comparative study of the main software of analysis, design and simulation of micro- network is realized. The components of our system have been modeled. Optimization methods and tools for graphical representation (Bond Graph, GIC and EMR) systems were presented. A survey and a literature review allowed us to evaluate the energy potential of the site and the needs of populations.- Chapter 3: This is the application of our tool on an isolated site identified in Senegal (MBoro / Mer). Optimization through the Annualized Cost of the System (ACS) objective function has allowed us to scale our system optimally. As the control system with Energetic Macroscopic Representation (EMR) has been designed.For further work should be considered a treatment of problems of availability of the system through monitoring and supervision of the system. The socio- economic component is also integrated in the future to meet the changing needs and habits of the population.

Este trabalho teve como objetivo avaliar a influência do ambiente competitivo na solução de problemas de localização de instalações e desenho da cadeia desenvolvendo um modelo matemático que captasse simultaneamente muitos dos aspectos práticos que impactam no funcionamento da cadeia de suprimentos. A seleção destes aspectos foi baseada nos comentários e sugestões que autores desta linha de pesquisa propuseram para aprofundamento da literatura da área. Foi desenvolvido um modelo de programação linear inteira mista (PLIM) aplicando-o em um problema exemplo para avaliar a hipótese de que a configuração de menor custo da cadeia não necessariamente proporciona maior acessibilidade ao mercado e, em seguida, o mesmo modelo foi aplicado a um problema real, de maior porte, no setor de fertilizantes. / The goal of this study was to evaluate the influence of a competitive environment in solving facility location and network design problems by developing a mathematical model that could capture several of the practical aspects that influence both decisions and operations of the supply chain. The selection of these aspects was based on the comments and suggestions proposed by several authors of this field of the literature. The mathematical program developed is based on mixed integer linear programming (MILP) and is firstly applied on an example problem to evaluate the hypothesis that the chain configuration that provides the lowest cost not necessarily also provides the greatest accessibility to the market. Afterwards, the same optimization model is applied to a larger real problem in the fertilizer industry.