Cette thèse aborde des problèmes liés à deux aspects majeurs de l'évolution d'Internet : l'aspect«comportemental», qui correspond aux nouvelles interactions entre les utilisateurs et le réseau, et l'aspect «structurel», lié aux changements d'Internet d'un point de vue architectural.Le manuscrit est composé d'un chapitre introductif qui donne les grandes lignes de recherche de ce travail de thèse, suivi d'un chapitre consacré à la description de l'état de l'art sur les deux aspects mentionnés ci-dessus. Parmi les solutions proposées par la communauté scientifique pour s'adapter à l'évolution d'Internet, deux nouveaux paradigmes réseaux sont particulièrement décrits : Information- Centric Networking (ICN) et Software-Defined Networking (SDN).La thèse continue avec la proposition de «Caesar», un dispositif réseau, inspiré par ICN, capable de gérer la distribution de contenus à partir de primitives de routage basées sur le nom des données et non les adresses des serveurs. Caesar est présenté dans deux chapitres, qui décrivent l'architecture et deux des principaux modules : le relayage et la gestion de la traçabilité des requêtes.La suite du manuscrit décrit un outil mathématique pour la détection efficace de boucles dans un réseau SDN d'un point de vue théorique. Les améliorations de l'algorithme proposé par rapport à l'état de l'art sont discutées. „,La thèse se conclue par un résumé des principaux résultats obtenus et une présentation des travaux en cours et futurs. / This thesis addresses two major aspects of the Internet evolution problem: a behavioral aspect, corresponding to a new type of interactions between users and the network, and a structural aspect, which refers to the evolution problem from an architectural point of view.The manuscript consists of an introductory chapter which outlines the research directions of this thesis, followed by a chapter on the description of the state of the art on the two aforementioned aspects.Among the solutions proposed by the scientific community to adapt to the evolution of the Internet, two new network paradigms are described: Information- Centric Networking (ICN) and Software-Defined Networking (SDN).The thesis continues with the description of "Caesar", a network device, inspired by ICN, capable of managing the distribution of content using forwarding primitives based on the content name and not a server address. Caesar is presented in two chapters describing the architecture of two main modules: theforwarding module, and the pending request management.The second part of the manuscript describes a mathematical tool for the effective loop detection in an SDN network from a theoretical point of view. Some algorithms are proposed and the improvements with respect ta the prior work are discussed.The thesis is concluded with a summary of the main results and a presentation of current and future work.

The rapid development in wireless technologies and multimedia services has radically shifted the major function of the current Internet from host-centric communication to service-oriented content dissemination, resulting a mismatch between the protocol design and the current usage patterns. Motivated by this significant change, Information-Centric Networking (ICN), which has been attracting ever-increasing attention from the communication networks research community, has emerged as a new clean-slate networking paradigm for future Internet. Through identifying and routing data by unified names, ICN aims at providing natural support for efficient information retrieval over the Internet. As a crucial characteristic of ICN, in-network caching enables users to efficiently access popular contents from on-path routers equipped with ubiquitous caches, leading to the enhancement of the service quality and reduction of network loads. Performance analysis and optimisation has been and continues to be key research interests of ICN. This thesis focuses on the development of efficient and accurate analytical models for the performance evaluation of ICN caching and the design of optimal caching management schemes under practical network configurations. This research starts with the proposition of a new analytical model for caching performance under the bursty multimedia traffic. The bursty characteristic is captured and the closed formulas for cache hit ratio are derived. To investigate the impact of topology and heterogeneous caching parameters on the performance, a comprehensive analytical model is developed to gain valuable insight into the caching performance with heterogeneous cache sizes, service intensity and content distribution under arbitrary topology. The accuracy of the proposed models is validated by comparing the analytical results with those obtained from extensive simulation experiments. The analytical models are then used as cost-efficient tools to investigate the key network and content parameters on the performance of caching in ICN. Bursty traffic and heterogeneous caching features have significant influence on the performance of ICN. Therefore, in order to obtain optimal performance results, a caching resource allocation scheme, which leverages the proposed model and targets at minimising the total traffic within the network and improving hit probability at the nodes, is proposed. The performance results reveal that the caching allocation scheme can achieve better caching performance and network resource utilisation than the default homogeneous and random caching allocation strategy. To attain a thorough understanding of the trade-off between the economic aspect and service quality, a cost-aware Quality-of-Service (QoS) optimisation caching mechanism is further designed aiming for cost-efficiency and QoS guarantee in ICN. A cost model is proposed to take into account installation and operation cost of ICN under a realistic ISP network scenario, and a QoS model is presented to formulate the service delay and delay jitter in the presence of heterogeneous service requirements and general probabilistic caching strategy. Numerical results show the effectiveness of the proposed mechanism in achieving better service quality and lower network cost. In this thesis, the proposed analytical models are used to efficiently and accurately evaluate the performance of ICN and investigate the key performance metrics. Leveraging the insights discovered by the analytical models, the proposed caching management schemes are able to optimise and enhance the performance of ICN. To widen the outcomes achieved in the thesis, several interesting yet challenging research directions are pointed out.

Ce travail illustre comment les tendances actuelles d'utilisation dominantes sur Internet motivent la recherche sur des architectures futures de réseau plus orientées vers le contenu. Parmi les architectures émergentes pour l'Internet du futur, le paradigme du Information-Centric Networking (ICN) est présenté. ICN vise à redéfinir les protocoles de base d'Internet afin d'y introduire un changement sémantique des hôtes vers les contenus. Parmi les architectures ICN, le Named-Data Networking (NDN) prévoit que les demandes nommées de contenus des utilisateurs soient transmises par leur nom dans les routeurs le long du chemin d'un consommateur à une ou plusieurs sources de contenus. Ces demandes de contenus laissent des traces dans les routeurs traversés qui sont ensuite suivis par les paquets de contenus demandés. La table d'intérêt en attente (PIT) est le composant du plan de données de l'NDN qui enregistre temporairement les demandes de contenus acheminés dans les routeurs. D'une part, ce travail explique que le mécanisme d'acheminement à états de la PIT permet des propriétés comme l'agrégation de requêtes, le multicast de réponses et le contrôle natif de flux hop-by-hop. D'autre part, ce travail illustre comment l'acheminement à états de la PIT peut facilement être mal utilisé par des attaquants pour monter des attaques de déni de service distribué (DDoS) disruptives, appelées Interest Flooding Attacks (IFAs). Dans les IFAs, des botnets vaguement coordonnés peuvent inonder le réseau d'une grande quantité de demandes difficiles à satisfaire dans le but de surcharger soit l'infrastructure du réseau soit les producteurs de contenus. Ce travail de thèse prouve que bien que des contre-mesures contre les IFAs aient été proposées, il manque une compréhension complète de leur efficacité réelle puisque celles-ci ont été testées sous des hypothèses simplistes sur les scénarios d'évaluation. Dans l'ensemble, le travail présenté dans ce manuscrit permet de mieux comprendre les implications des IFAs et les opportunités d'améliorer les mécanismes de défense existants contre ces attaques. Les principales contributions de ce travail de thèse tournent autour d'une analyse de sécurité du plan d'acheminement dans l'architecture NDN. En particulier, ce travail définit un modèle d'attaquant plus robuste pour les IFAs à travers l'identification des failles dans les contre-mesures IFA existantes. Ce travail introduit un nouvel ensemble d'IFAs basé sur le modèle d'attaquant proposé. Les nouveaux IFAs sont utilisés pour réévaluer les plus efficaces contre-mesures IFA existantes. Les résultats de cette évaluation réfutent l'efficacité universelle des mécanismes de défense existants contre l'IFA et, par conséquent, appellent à différentes contre-mesures pour protéger le NDN contre cette menace de sécurité. Pour surmonter le problème révélé, ce travail définit également des contre-mesures proactives contre l'IFA, qui sont de nouveaux mécanismes de défense contre les IFA inspirés par les problèmes rencontrés dans l'état de l'art. Ce travail présente Charon, une nouvelle contre-mesure proactive contre l'IFA, et la teste contre les nouvelles IFAs. Ce travail montre que Charon est plus efficace que les contre-mesures IFA réactives existantes. Enfin, ce travail illustre la conception NDN.p4, c'est-à-dire la première implémentation d'un protocole ICN écrit dans le langage de haut niveau pour les processeurs de paquets P4. Le travail NDN.p4 est la première tentative dans la littérature visant à tirer parti des nouveaux techniques de réseaux programmables pour tester et évaluer différentes conceptions de plan de données NDN. Cette dernière contribution classe également les mécanismes alternatifs d'acheminement par rapport à un ensemble de propriétés cardinales de la PIT. Le travail souligne qu'il vaut la peine d'explorer d'autres mécanismes d'acheminement visant à concevoir un plan de données NDN moins vulnérable à la menace IFA / This work illustrates how today's Internet dominant usage trends motivate research on more content-oriented future network architectures. Among the emerging future Internet proposals, the promising Information-Centric Networking (ICN) research paradigm is presented. ICN aims to redesign Internet's core protocols to promote a shift in focus from hosts to contents. Among the ICN architectures, the Named-Data Networking (NDN) envisions users' named content requests to be forwarded by their names in routers along the path from one consumer to 1-or-many sources. NDN's requests leave trails in traversed routers which are then followed backwards by the requested contents. The Pending Interest Table (PIT) is the NDN's data-plane component which temporarily records forwarded content requests in routers. On one hand, this work explains that the PIT stateful mechanism enables properties like requests aggregation, multicast responses delivery and native hop-by-hop control flow. On the other hand, this work illustrates how the PIT stateful forwarding behavior can be easily abused by malicious users to mount disruptive distributed denial of service attacks (DDoS), named Interest Flooding Attacks (IFAs). In IFAs, loosely coordinated botnets can flood the network with a large amount of hard to satisfy requests with the aim to overload both the network infrastructure and the content producers. This work proves that although countermeasures against IFAs have been proposed, a fair understanding of their real efficacy is missing since those have been tested under simplistic assumptions about the evaluation scenarios. Overall, the work presented in this manuscript shapes a better understanding of both the implications of IFAs and the possibilities of improving the state-of-the-art defense mechanisms against these attacks. The main contributions of this work revolves around a security analysis of the NDN's forwarding plane. In particular, this work defines a more robust attacker model for IFAs by identifying flaws in the state-of-the-art IFA countermeasures. This work introduces a new set of IFAs built upon the proposed attacker model. The novel IFAs are used to re-assess the most effective existing IFA countermeasures. Results of this evaluation disproves the universal efficacy of the state-of-the-art IFA defense mechanisms and so, call for different countermeasures to protect the NDN against this threat. To overcome the revealed issue, this work also defines proactive IFA countermeasures, which are novel defense mechanisms against IFAs inspired by the issues with the state-of-the-art ones. This work introduces Charon, a novel proactive IFA countermeasure, and tests it against the novel IFA attacks. This work shows Charon counteracts latest stealthy IFAs better than the state-of-the-art reactive countermeasures. Finally, this work illustrates the NDN.p4 design, that is, the first implementation of an ICN protocol written in the high-level language for packet processors P4. The NDN.p4 work is the first attempt in the related literature to leverage novel programmable-networks technologies to test and evaluate different NDN forwarding plane designs. This last contribution also classifies existing alternative forwarding mechanisms with respect to a set of PIT cardinal properties. The work outlines that it is worth to explore alternative forwarding mechanisms aiming to design an NDN forwarding plane less vulnerable to the IFA threat

L’Internet a été initialement conçu pour servir des usages fixes et sédentaires, cependant les projections montrent que les futurs utilisateurs d'Internet seront de plus en plus mobiles. A l'heure actuelle, la rapidité avec laquelle cette évolution se déroule et la gestion souvent insuffisante des réseaux d'accès représentent un obstacle majeur au développement de services avancés. Afin de résoudre ces problèmes et répondre aux besoins de l'Internet mobile, les fournisseurs de services ont besoin de maîtriser l'expansion de la capacité nécessaire dans leurs réseaux de collecte, sinon le trafic de donnés va pouvoir boucher leurs réseaux dans le futur. Le déploiement des nouvelles générations de réseaux fournit des hautes bandes passantes et débits mais implique souvent des grandes dépenses en capital et en exploitation. Une alternative économiquement et techniquement viable est représentée par les solutions de déchargement du trafic mobile. Ces solutions peuvent réduire la surcharge sur le spectre radio et sur les stations de base et sur le réseau de collecte. Les solutions de déchargement les plus couramment utilisées sont le déchargement sur les réseaux de femtocellules et les réseaux Wi-Fi. Dans le même contexte, pour résoudre le problème de congestion dans le réseau cellulaire, une nouvelle solution est récemment apparue: Information Centric networking permettant la mise en cache des contenus dans le réseau ce qui minimise le temps d'accès aux contenus. L'objectif de cette thèse est donc d'étudier ces nouvelles solutions de déchargement de trafic et de contenu dans les réseaux cellulaires en prenant en considération les schémas de mobilité et les comportements humains. / The Internet was initially conceived to serve fix and sedentary usages, while current socio-technological trends clearly show that future Internet users will be increasingly mobile and nomadic. At present, the speed at which this evolution takes place and the inadequate management of access networks represent a major obstacle in the development of advanced services. To solve these problems and to meet the needs of mobile Internet, service providers need to master the needed capacity expansion in their backhauling network, otherwise the data traffic will clog their networks in the future. Next-generation network deployments promise to deliver higher bandwidth and speed, but they often imply high capital and operational expenditures. An alternative economically and technically viable way is represented by mobile data offloading solutions. These solutions can reduce the load on radio spectrum, on base stations and on backhauling network. The most commonly used offloading solutions are over small-networks and over Wi-Fi networks. In the same context, and in order to solve the problem of congestion in the cellular network, a new solution has emerged recently : Information Centric Networking for in-network caching that permits to minimize content access latency. The objective of this thesis is to study these new traffic and content offloading solutions in cellular networks while taking into account the mobility patterns and human behavior.

Um desafio fundamental em ICN (do inglês Information-Centric Networking) é desenvolver Content Stores (ou seja, unidades de cache) que satisfaçam três requisitos: espaço de armazenamento grande, velocidade de operação rápida e custo acessível. A chamada Hierarchical Content Store (HCS) é uma abordagem promissora para atender a esses requisitos. Ela explora a correlação temporal entre requisições para prever futuras solicitações. Por exemplo, assume-se que um usuário que solicita o primeiro minuto de um filme também solicitará o segundo minuto. Teoricamente, essa premissa permitiria transferir proativamente conteúdos de uma área de cache relativamente grande, mas lenta (Layer 2 - L2), para uma área de cache mais rápida, porém menor (Layer 1 - L1). A estrutura hierárquica tem potencial para incrementar o desempenho da CS em uma ordem de grandeza tanto em termos de vazão como de tamanho, mantendo o custo. Contudo, o desenvolvimento de HCS apresenta diversos desafios práticos. É necessário acoplar as hierarquias de memória L2 e L1 considerando as suas taxas de transferência e tamanhos, que dependem tanto de aspectos de hardware (por exemplo, taxa de leitura da L2, uso de múltiplos SSD físicos em paralelo, velocidade de barramento, etc.), como de software (por exemplo, controlador do SSD, gerenciamento de memória, etc.). Nesse contexto, esta tese apresenta duas contribuições principais. Primeiramente, é proposta uma arquitetura para superar os gargalos inerentes ao sistema através da paralelização de múltiplas HCS. Em resumo, o esquema proposto supera desafios inerentes à concorrência (especificamente, sincronismo) através do particionamento determinístico das requisições de conteúdos entre múltiplas threads. Em segundo lugar, é proposta uma metodologia para investigar o desenvolvimento de HCS explorando técnicas de emulação e modelagem analítica conjuntamente. A metodologia proposta apresenta vantagens em relação a metodologias baseadas em prototipação e simulação. A L2 é emulada para viabilizar a investigação de uma variedade de cenários de contorno (tanto em termos de hardware como de software) maior do que seria possível através de prototipação (considerando as tecnologias atuais). Além disso, a emulação emprega código real de um protótipo para os outros componentes do HCS (por exemplo L1, gerência das camadas e API) para fornecer resultados mais realistas do que seriam obtidos através de simulação. / A key challenge in Information Centric Networking (ICN) is to develop cache units (also called Content Store - CS) that meet three requirements: large storage space, fast operation, and affordable cost. The so-called HCS (Hierarchical Content Store) is a promising approach to satisfy these requirements jointly. It explores the correlation between content requests to predict future demands. Theoretically, this idea would enable proactively content transfers from a relatively large but slow cache area (Layer 2 - L2) to a faster but smaller cache area (Layer 1 - L1). Thereby, it would be possible to increase the throughput and size of CS in one order of magnitude, while keeping the cost. However, the development of HCS introduces several practical challenges. HCS requires a careful coupling of L2 and L1 memory levels considering their transfer rates and sizes. This requirement depends on both hardware specifications (e.g., read rate L2, use of multiple physical SSD in parallel, bus speed, etc.), and software aspects (e.g., the SSD controller, memory management, etc.). In this context, this thesis presents two main contributions. First, we propose an architecture for overcoming the HCS bottlenecks by parallelizing multiple HCS. In summary, the proposed scheme overcomes racing condition related challenges through deterministic partitioning of content requests among multiple threads. Second, we propose a methodology to investigate the development of HCS exploiting emulation techniques and analytical modeling jointly. The proposed methodology offers advantages over prototyping and simulation-based methods. We emulate the L2 to enable the investigation of a variety of boundary scenarios that are richer (regarding both hardware and software aspects) than would be possible through prototyping (considering current technologies). Moreover, the emulation employs real code from a prototype for the other components of the HCS (e.g., L1, layers management and API) to provide more realistic results than would be obtained through simulation.

As the Internet is becoming information-centric, network services increasingly demand scalable and efficient communication of information between a multitude of information producers and large groups of interested information consumers. Such information-centric services are growing rapidly in use and deployment. Examples of deployed services that are information-centric include: IPTV, MMORPG, VoD, video conferencing, file sharing, software updates, RSS dissemination, online markets, and grid computing. To effectively support future information-centric services, the network infrastructure for multi-point communication has to address a number of significant challenges: (i) how to understand massive information-centric groups in a scalable manner, (ii) how to analyze and predict the evolution of those groups in an accurate and efficient way, and (iii) how to disseminate content from information producers to a vast number of groups with potentially long-lived membership and highly diverse, dynamic group activity levels? This dissertation proposes novel architecture and protocols that effectively address the above challenges in supporting multi-point communication for future information-centric network services. In doing so, we make the following three major contributions: (1) We develop a novel technique called Proximity Embedding (PE) that can approximate a family of path-ensembled based proximity measures for information-centric groups. We develop Clustered Spectral Graph Embedding (SCGE) that captures the essential structure of large graphs in a highly efficient and scalable manner. Our techniques help to explain the proximity (closeness) of users in information-centric groups, and can be applied to a variety of analysis tasks of complex network structures. (2) Based on SCGE, we develop new supervision based link prediction techniques called Clustered Spectral Learning and Clustered Polynomial Learning that enable us to predict the evolution of massive and complex network structures in an accurate and efficient way. By exploiting supervised information from past snapshots of network structures, our methods yield up to 20% improvement in link prediction accuracy when compared to existing state-of-the-art methods. (3) Finally, we develop a novel multicast infrastructure called Multicast with Adaptive Dual-state (MAD). MAD supports large number of group and group membership, and efficient content dissemination in a presence of dynamic group activity. We demonstrate the effectiveness of our approach in extensive simulation, analysis, and emulation through the real system implementation. / text

Evolving smart infrastructures requires both content distribution as well as event notification and processing support. Content Centric Networking (CCN), built around named data, is a clean slate network architecture for supporting future applications. Due to its focus on content distribution, CCN does not inherently support Publish-Subscribe event notification, a fundamental building block in computer mediated systems and a critical requirement for smart infrastructure applications. While semantics of content distribution and event notification require different support systems from the underlying network infrastructure, content distribution and event notification can still be united by leveraging similarities in the routing infrastructure. Our Extended-CCN architecture(X-CCN) realizes this to provide lightweight content based pub-sub service at the network layer, which is used to provide advanced publish/subscribe services at higher layers. Light weight content based pub-sub and CCN communication at network layer along with advanced publish/subscribe together are presented as data fabric for the smart infrastructures applications.

Evolving smart infrastructures requires both content distribution as well as event notification and processing support. Content Centric Networking (CCN), built around named data, is a clean slate network architecture for supporting future applications. Due to its focus on content distribution, CCN does not inherently support Publish-Subscribe event notification, a fundamental building block in computer mediated systems and a critical requirement for smart infrastructure applications. While semantics of content distribution and event notification require different support systems from the underlying network infrastructure, content distribution and event notification can still be united by leveraging similarities in the routing infrastructure. Our Extended-CCN architecture(X-CCN) realizes this to provide lightweight content based pub-sub service at the network layer, which is used to provide advanced publish/subscribe services at higher layers. Light weight content based pub-sub and CCN communication at network layer along with advanced publish/subscribe together are presented as data fabric for the smart infrastructures applications.

Um desafio fundamental em ICN (do inglês Information-Centric Networking) é desenvolver Content Stores (ou seja, unidades de cache) que satisfaçam três requisitos: espaço de armazenamento grande, velocidade de operação rápida e custo acessível. A chamada Hierarchical Content Store (HCS) é uma abordagem promissora para atender a esses requisitos. Ela explora a correlação temporal entre requisições para prever futuras solicitações. Por exemplo, assume-se que um usuário que solicita o primeiro minuto de um filme também solicitará o segundo minuto. Teoricamente, essa premissa permitiria transferir proativamente conteúdos de uma área de cache relativamente grande, mas lenta (Layer 2 - L2), para uma área de cache mais rápida, porém menor (Layer 1 - L1). A estrutura hierárquica tem potencial para incrementar o desempenho da CS em uma ordem de grandeza tanto em termos de vazão como de tamanho, mantendo o custo. Contudo, o desenvolvimento de HCS apresenta diversos desafios práticos. É necessário acoplar as hierarquias de memória L2 e L1 considerando as suas taxas de transferência e tamanhos, que dependem tanto de aspectos de hardware (por exemplo, taxa de leitura da L2, uso de múltiplos SSD físicos em paralelo, velocidade de barramento, etc.), como de software (por exemplo, controlador do SSD, gerenciamento de memória, etc.). Nesse contexto, esta tese apresenta duas contribuições principais. Primeiramente, é proposta uma arquitetura para superar os gargalos inerentes ao sistema através da paralelização de múltiplas HCS. Em resumo, o esquema proposto supera desafios inerentes à concorrência (especificamente, sincronismo) através do particionamento determinístico das requisições de conteúdos entre múltiplas threads. Em segundo lugar, é proposta uma metodologia para investigar o desenvolvimento de HCS explorando técnicas de emulação e modelagem analítica conjuntamente. A metodologia proposta apresenta vantagens em relação a metodologias baseadas em prototipação e simulação. A L2 é emulada para viabilizar a investigação de uma variedade de cenários de contorno (tanto em termos de hardware como de software) maior do que seria possível através de prototipação (considerando as tecnologias atuais). Além disso, a emulação emprega código real de um protótipo para os outros componentes do HCS (por exemplo L1, gerência das camadas e API) para fornecer resultados mais realistas do que seriam obtidos através de simulação. / A key challenge in Information Centric Networking (ICN) is to develop cache units (also called Content Store - CS) that meet three requirements: large storage space, fast operation, and affordable cost. The so-called HCS (Hierarchical Content Store) is a promising approach to satisfy these requirements jointly. It explores the correlation between content requests to predict future demands. Theoretically, this idea would enable proactively content transfers from a relatively large but slow cache area (Layer 2 - L2) to a faster but smaller cache area (Layer 1 - L1). Thereby, it would be possible to increase the throughput and size of CS in one order of magnitude, while keeping the cost. However, the development of HCS introduces several practical challenges. HCS requires a careful coupling of L2 and L1 memory levels considering their transfer rates and sizes. This requirement depends on both hardware specifications (e.g., read rate L2, use of multiple physical SSD in parallel, bus speed, etc.), and software aspects (e.g., the SSD controller, memory management, etc.). In this context, this thesis presents two main contributions. First, we propose an architecture for overcoming the HCS bottlenecks by parallelizing multiple HCS. In summary, the proposed scheme overcomes racing condition related challenges through deterministic partitioning of content requests among multiple threads. Second, we propose a methodology to investigate the development of HCS exploiting emulation techniques and analytical modeling jointly. The proposed methodology offers advantages over prototyping and simulation-based methods. We emulate the L2 to enable the investigation of a variety of boundary scenarios that are richer (regarding both hardware and software aspects) than would be possible through prototyping (considering current technologies). Moreover, the emulation employs real code from a prototype for the other components of the HCS (e.g., L1, layers management and API) to provide more realistic results than would be obtained through simulation.

Um desafio fundamental em ICN (do inglês Information-Centric Networking) é desenvolver Content Stores (ou seja, unidades de cache) que satisfaçam três requisitos: espaço de armazenamento grande, velocidade de operação rápida e custo acessível. A chamada Hierarchical Content Store (HCS) é uma abordagem promissora para atender a esses requisitos. Ela explora a correlação temporal entre requisições para prever futuras solicitações. Por exemplo, assume-se que um usuário que solicita o primeiro minuto de um filme também solicitará o segundo minuto. Teoricamente, essa premissa permitiria transferir proativamente conteúdos de uma área de cache relativamente grande, mas lenta (Layer 2 - L2), para uma área de cache mais rápida, porém menor (Layer 1 - L1). A estrutura hierárquica tem potencial para incrementar o desempenho da CS em uma ordem de grandeza tanto em termos de vazão como de tamanho, mantendo o custo. Contudo, o desenvolvimento de HCS apresenta diversos desafios práticos. É necessário acoplar as hierarquias de memória L2 e L1 considerando as suas taxas de transferência e tamanhos, que dependem tanto de aspectos de hardware (por exemplo, taxa de leitura da L2, uso de múltiplos SSD físicos em paralelo, velocidade de barramento, etc.), como de software (por exemplo, controlador do SSD, gerenciamento de memória, etc.). Nesse contexto, esta tese apresenta duas contribuições principais. Primeiramente, é proposta uma arquitetura para superar os gargalos inerentes ao sistema através da paralelização de múltiplas HCS. Em resumo, o esquema proposto supera desafios inerentes à concorrência (especificamente, sincronismo) através do particionamento determinístico das requisições de conteúdos entre múltiplas threads. Em segundo lugar, é proposta uma metodologia para investigar o desenvolvimento de HCS explorando técnicas de emulação e modelagem analítica conjuntamente. A metodologia proposta apresenta vantagens em relação a metodologias baseadas em prototipação e simulação. A L2 é emulada para viabilizar a investigação de uma variedade de cenários de contorno (tanto em termos de hardware como de software) maior do que seria possível através de prototipação (considerando as tecnologias atuais). Além disso, a emulação emprega código real de um protótipo para os outros componentes do HCS (por exemplo L1, gerência das camadas e API) para fornecer resultados mais realistas do que seriam obtidos através de simulação. / A key challenge in Information Centric Networking (ICN) is to develop cache units (also called Content Store - CS) that meet three requirements: large storage space, fast operation, and affordable cost. The so-called HCS (Hierarchical Content Store) is a promising approach to satisfy these requirements jointly. It explores the correlation between content requests to predict future demands. Theoretically, this idea would enable proactively content transfers from a relatively large but slow cache area (Layer 2 - L2) to a faster but smaller cache area (Layer 1 - L1). Thereby, it would be possible to increase the throughput and size of CS in one order of magnitude, while keeping the cost. However, the development of HCS introduces several practical challenges. HCS requires a careful coupling of L2 and L1 memory levels considering their transfer rates and sizes. This requirement depends on both hardware specifications (e.g., read rate L2, use of multiple physical SSD in parallel, bus speed, etc.), and software aspects (e.g., the SSD controller, memory management, etc.). In this context, this thesis presents two main contributions. First, we propose an architecture for overcoming the HCS bottlenecks by parallelizing multiple HCS. In summary, the proposed scheme overcomes racing condition related challenges through deterministic partitioning of content requests among multiple threads. Second, we propose a methodology to investigate the development of HCS exploiting emulation techniques and analytical modeling jointly. The proposed methodology offers advantages over prototyping and simulation-based methods. We emulate the L2 to enable the investigation of a variety of boundary scenarios that are richer (regarding both hardware and software aspects) than would be possible through prototyping (considering current technologies). Moreover, the emulation employs real code from a prototype for the other components of the HCS (e.g., L1, layers management and API) to provide more realistic results than would be obtained through simulation.