Search

[ANGLÈS] Software-Defined Networking (SDN) has been recognized as the next-generation networking paradigm. It is a fast-evolving technology that decouples the network data plane, which are the network devices that forward traffic, from the network control plane, which is the software logic that controls ultimately how traffic is forwarded through the network. A logically centralized controller is responsible for all the control decisions and communication among the forwarding elements. It allows to control the network behavior from a single high level control program. However, Software-Defined Networks use many of its network resources inefficiently, which leads to over-loading network links, congestion in the queues and end-to-end packet delays. Consequently, it becomes clear that routing decisions affect the overall performance of a communication's network. Performance is determined in terms of Quality of Service guarantees, i.e. throughput, average packet delay, jitter and losses. Thus, a QoS-aware routing algorithm is required. Current traffic engineering techniques and state-of-the-art routing algorithms do not effectively use the merits of SDNs, such as global centralized visibility, real-time fast decisions, control and data plane decoupling, network management simplification and portability. In this thesis, we developed two new QoS-aware routing algorithms that exploit the advantages that SDN brings to improve the network performance. Two different scenarios have been studied: a centralized and a distributed models. The centralized scenario simplifies the management of complex flows and the customization, but scalability issues arise. In contrast, the distributed scenario is more scalable, but there may be state inconsistency and increase of shared information. In general, a centralized approach is better for data centers or home networks, whereas a distributed approach is better for large scale networks, e.g. cloud environments. First, the centralized SDN controller model is discussed, for which a multi-tenancy management framework is proposed to fulfill the quality-of-services (QoSs) requirements through tenant isolation, prioritization and flow allocation. A network virtualization algorithm is provided to isolate and prioritize tenants from different clients. Furthermore, a novel routing scheme, called QoS-aware Virtualization-enabled Routing (QVR), is presented. It combines the proposed virtualization technique and a QoS-aware framework to enable flow allocation with respect to different tenant applications. Simulation results confirm that the proposed QVR algorithm surpasses the conventional algorithms with less traffic congestion and packet delay. This facilitates reliable and efficient data transportation in generalized SDNs. Therefore, it yields to service performance improvement for numerous applications and enhancement of client isolation. Second, a distributed SDN controller model is analyzed. The network is divided into different clusters, and hierarchically split in two levels. This architecture leads to smaller sizes of routing tables in the switches, and substantially lesser calculations and updates of routing tables from the controller. Moreover, a new algorithm is developed, called QoS-aware Reinforcement Learning Routing (QRLR), where reinforcement learning is applied to the routing problem. The modeling of the reward function calculation solution allows the customization of the different requirements for each type of traffic, thus providing flexibility and adaptability to different flows and its requirements. [CASTELLÀ] Las redes definidas por software (SDN) han sido reconocidas como el paradigma de las redes de nueva generación. Es una tecnología de rápida evolución que desacopla el plano de datos de la red, que son los dispositivos de la red que transmiten el tráfico, desde el plano de control de la red, que es la lógica de software que controla en última instancia cómo el tráfico se transmite a través de la red. En SDN, un controlador centralizado es responsable de todas las decisiones de control y la comunicación entre los elementos de la red. Permite controlar el comportamiento de la red desde un único punto de control de alto nivel. Sin embargo, las redes definidas por software utilizan muchos de sus recursos de manera ineficiente, lo que conduce a un exceso de carga de los enlaces de la red, congestión en las colas y retrasos en el envió de los paquetes. En consecuencia, se hace evidente que las decisiones de enrutamiento afectan el rendimiento global de la red de una comunicación. El rendimiento se determina en función de la calidad de servicio (QoS), es decir, el rendimiento, la demora media de paquetes, el jitter y las pérdidas de paquetes. Por lo tanto, se requiere un algoritmo de encaminamiento con QoS. Las técnicas de ingeniería de tráfico actuales y algoritmos de enrutamiento no utilizan eficazmente los méritos de SDN, tales como la visibilidad global centralizada, las decisiones rápidas en tiempo real, el desacoplamiento de los planos de control y de datos, la simplificación de la gestión de redes y la portabilidad. En esta tesis, hemos desarrollado dos nuevos algoritmos de encaminamiento con QoS que explotan las ventajas que trae SDN para mejorar el rendimiento de la red. Dos escenarios diferentes se han estudiado: un modelo centralizado y un modelo distribuido. El escenario centralizado simplifica la gestión de los flujos complejos y la personalización, pero surgen problemas de escalabilidad. En contraste, el escenario distribuido es más escalable, pero puede haber inconsistencia en los estados, y la información compartida entre los elementos de la red aumenta. En general, un enfoque centralizado es mejor para centros de datos o redes domésticas, mientras que un enfoque distribuido es mejor para las redes a gran escala, por ejemplo, entornos de la nube. En primer lugar, se analiza el modelo de controlador SDN centralizado, para el que se propone una gestión para cumplir los requisitos de QoS a través de aislamiento, priorización y enrutamiento. Un algoritmo de virtualización de la red se proporciona para aislar y dar prioridad a los datos de los diferentes clientes. Resultados de la simulación confirman que el algoritmo propuesto supera los algoritmos convencionales con menor congestión de tráfico y retardo de los paquetes. En segundo lugar, se analiza un modelo de controlador SDN distribuido. La red se divide en diferentes grupos, y jerárquicamente se dividide en dos niveles. Esta arquitectura conduce a tamaños más pequeños de las tablas de enrutamiento, cálculos sustancialmente menores y actualizaciones de las tablas de enrutamiento desde el controlador. Por otra parte, el aprendizaje por refuerzo se aplica al problema de enrutamiento. El modelado de la función recompensa permite la personalización de los diferentes requisitos para cada tipo de tráfico, lo que proporciona flexibilidad y capacidad de adaptación a los diferentes flujos y sus requisitos. [CATALÀ] Les xarxes definides per software (SDN) han estat reconegudes com el paradigma de les xarxes de nova generació. És una tecnologia de ràpida evolució que desacobla el pla de dades de la xarxa, que són els dispositius de la xarxa que transmeten els paquets, del pla de control de la xarxa, que és la lògica que controla en última instància com el trànsit es transmet a través de la xarxa. En SDN, un controlador centralitzat és responsable de totes les decisions de control i la comunicació entre els elements de la xarxa. Permet controlar el comportament de la xarxa des d'un únic punt de control d'alt nivell. No obstant això, les xarxes definides per software utilitzen molts dels seus recursos de manera ineficient, el que condueix a un excés de càrrega dels enllaços de la xarxa, congestió en les cues i retards en la transmissió dels paquets. En conseqüència, es fa evident que les decisions d'encaminament que afecten el rendiment global de la xarxa han de ser estudiades. El rendiment es determina en funció de la qualitat de servei (QoS), és a dir, el rendiment, la demora mitjana de paquets, el jitter i les pèrdues de paquets. Per tant, es requereix d'un algoritme d'encaminament amb QoS. Les tècniques d'enginyeria de trànsit actuals i algoritmes d'encaminament no utilitzen eficaçment els mèrits de SDN, com ara la visibilitat global centralitzada, les decisions ràpides en temps real, el desacoblament dels plans de control i de dades, la simplificació de la gestió de les xarxes i la portabilitat. En aquesta tesi, hem desenvolupat dos nous algoritmes d'encaminament amb QoS que exploten els avantatges que porta SDN per millorar el rendiment de la xarxa. S'han estudiat dos escenaris diferents: un model centralitzat i un model distribuït. L'escenari centralitzat simplifica la gestió dels fluxos complexos i la personalització, però sorgeixen problemes d'escalabilitat. En contrast, l'escenari distribuït és més escalable, però pot ocasionar inconsistència en els estats, i la informació compartida entre els elements de la xarxa augmenta. En general, un enfocament centralitzat és millor per a centres de dades o xarxes domèstiques, mentre que un enfocament distribuït és millor per a les xarxes a gran escala, per exemple, entorns del núvol. En primer lloc, s'analitza el model de controlador SDN centralitzat, per al qual es proposa una gestió per complir els requisits de QoS a través d'aïllament, priorització i enrutament. Es presenta un algoritme de virtualització de la xarxa per aïllar i donar prioritat a les dades dels diferents clients. Els resultats de la simulació confirmen que l'algoritme proposat supera els algoritmes convencionals amb menor congestió de trànsit i menys retard dels paquets. En segon lloc, s'analitza un model de controlador SDN distribuït. La xarxa es divideix en diferents grups, i jeràrquicament es divideix en dos nivells. Aquesta arquitectura condueix a una reducció de les taules d'enrutament, càlculs substancialment menors i actualitzacions de les taules d'encaminament des de diferents controladors. A més, s'aplica l'aprenentatge per reforç al problema d'enrutament. El modelatge de la funció de recompensa permet la personalització dels diferents requisits per a cada tipus de trànsit, el que proporciona flexibilitat i capacitat d'adaptació als diferents fluxos i els seus requisits.

The growth of the Internet in terms of number of devices, the number of networks associated to each device and the mobility of devices and users makes the operation and management of the Internet network infrastructure a very complex challenge. In order to address this challenge, innovative solutions and ideas must be tested and evaluated in real network environments and not only based on simulations or laboratory setups. OFELIA is an European FP7 project and its main objective is to address the aforementioned challenge by building and operating a multi-layer, multi-technology and geographically distributed Future Internet testbed facility, where the network itself is precisely controlled and programmed by the experimenter using the emerging OpenFlow technology. This paper reports on the work done during the first half of the project, the lessons learned as well as the key advantages of the OFELIA facility for developing and testing new networking ideas. An overview on the challenges that have been faced on the design and implementation of the testbed facility is described, including the OFELIA Control Framework testbed management software. In addition, early operational experience of the facility since it was opened to the general public, providing five different testbeds or islands, is described.

The number of devices connected to the Internet is increasing, data centers are growing continuously and computer networks are getting more complex. Traditional network management approach is becoming more difficult and insufficient. Software-Defined Networks (SDN) is a new generation networking approach which is expected to take place of the traditional computer networks. SDN architecture provides effective management of the large and complex networks. Although SDN have benefits from the network security perspective, it also brings new attack vectors. We believe that the network security problems in SDN architecture need more advanced solutions. In this work, a survey on the SDN security problems is presented, challenges are discussed. In this context, security threats and attack surfaces in SDN are described, the significant SDN security solution examples in the literature are given.

Made available in DSpace on 2015-04-29T15:10:48Z (GMT). No. of bitstreams: 1 ALEXANDRE PASSITO.pdf: 3822416 bytes, checksum: 4f278e2830ed590e916983c979c90872 (MD5) Previous issue date: 2012-12-04 CAPES - Coordena????o de Aperfei??oamento de Pessoal de N??vel Superior Autonomous networking research to applies intelligent agent and multiagent systems theory to network controlling mechanisms. Deploying such autonomous and rational entities in the network can improve its behavior in the presence of very dynamic and complex control scenarios. Unfortunately, building agent-based mechanisms for networks is not an easy task. The main difficulty is to create concise knowledge representations about network domains and reasoning mechanisms to deal with them. Furthermore, the Internet makes the design of multiagent systems for network controlling a challenging activity involving the modeling of different participants with diverse beliefs and intentions. Such type of system often poses scalability problems due to the lack of incentives for cooperation between administrative domains. Finally, as the current structure of the Internet often prevents innovation, constructed autonomous networking mechanisms are not fully deployed in large scale scenarios. The Software-Defined Networking (SDN) paradigm is in the realm of Future Internet efforts. In the SDN paradigm, packet forwarding hardware is controlled by software running as a separated control plane. Management software uses an open protocol to program the owtables in different switches and routers. This work presents a general discussion about the integration of autonomous networks and software-defined networks. Based on the knowledge offered by this discussion, it presents a framework that provides autonomy to SDN domains allowing them to act cooperatively when deployed in scenarios with distributed management. Two case studies are presented for important open issues in the Internet: (1) the problem of mitigating DDoS attacks when thousands of attackers perform malicious packet ooding and SDN domains must cooperate to cope with packet filtering at the source; (2) the problem of network traffic management when multiple domains must cooperate and modify routing primitives. A pesquisa em redes aut??nomas aplica a teoria de agentes inteligentes e sistemas multiagente em mecanismos de controle de redes. Implantar esse mecanismos aut??nomos e racionais na rede pode melhorar seu comportamento na presen??a de cen??rios de controle muito complexos e din??micos. Infelizmente, a constru????o de mecanismos baseados em agentes para redes n??o ?? uma tarefa f??cil. A principal dificuldade ?? criar representa????es concisas de conhecimento sobre os dom??nios de redes e mecanismos de racioc??nio para lidar com elas. Al??m disso, a Internet faz com que o projeto de sistemas multiagente para o controle da rede seja uma atividade intrincada envolvendo a modelagem de diferentes participantes com diversas cren??as e inten????es. Esses tipos de sistemas geralmente apresentam problemas de escalabilidade devido ?? falta de incentivos para coopera????o entre dom??nios administrativos. Finalmente, como a estrutura corrente da Internet geralmente impede inova????es, mecanismos de redes aut??nomas constru??dos n??o s??o totalmente implantados em cen??rios de larga escala. O paradigma das redes definidas por software (SDN) est?? na esfera dos esfor??os da Internet do Futuro. No paradigma SDN, o hardware de repasse de pacotes ?? controlado por software sendo executado como um plano de controle separado. Softwares de gerenciamento utilizam um protocolo aberto que programa as tabelas de uxo em diferentes switches e roteadores. Este trabalho apresenta uma discuss??o geral sobre a integra????o de redes aut??nomas e redes definidas por software. Baseado no conhecimento oferecido por essa discuss??o, ?? apresentado um arcabou??o que prov?? autonomia para dom??nios SDN, permitindo que eles atuem cooperativamente quando implantados em cen??rios com gerenciamento distribu??do. Dois estudos de caso s??o apresentados para importantes quest??es em aberto na Internet: (1) o problema da mitiga????o de ataques DDoS quando milhares de atacantes realizam inunda????o por pacotes e os dom??nios SDN precisam cooperar para lidar com o filtro de pacotes na origem; (2) o problema do gerenciamento de tr??fego da rede quando m??ltiplos dom??nios devem cooperar e realizar modifica????es nas primitivas de roteamento de redes.

The growing energy consumption of communication networks has attracted the attention of the networking researchers in the last decade. In this context, the new architecture of Software-Defined Networks (SDN) allows a flexible programmability, suitable for the power-consumption optimization problem. In this paper we address the issue of designing a novel distributed routing algorithm that optimizes the power consumption in large scale SDN with multiple domains. The solution proposed, called DEAR (Distributed Energy- Aware Routing), tackles the problem of minimizing the number of links that can be used to satisfy a given data traffic demand under performance constraints such as control traffic delay and link utilization. To this end, we present a complete formulation of the optimization problem that considers routing requirements for control and data plane communications. Simulation results confirm that the proposed solution enables the achievement of significant energy savings.

In OpenFlow-based Software-Defined Networks, obtaining flow-level measurements, similar to those provided by NetFlow/IPFIX, is challenging as it requires to install an entry per flow in the flow tables. This approach does not scale well as the number of entries in the flow tables is limited and small. Moreover, labeling the flows with the application that generates the traffic would greatly enrich these reports, as it would provide very valuable information for network performance and security among others. In this paper, we present a scalable flow monitoring solution fully compatible with current off-the-shelf OpenFlow switches. Measurements are maintained in the switches and are asynchronously sent to a SDN controller. Additionally, flows are classified using a combination of DPI and Machine Learning (ML) techniques with special focus on the identification of web and encrypted traffic. For the sake of scalability, we designed two different traffic sampling methods depending on the OpenFlow features available in the switches. We implemented our monitoring solution within OpenDaylight and evaluated it in a testbed with Open vSwitch, using also a number of DPI and ML tools to find the best tradeoff between accuracy and performance. Our experimental results using real-world traffic show that the measurement and classification systems are accurate and the cost to deploy them is significantly reduced.

Нарастващите информационни потребности в корпоративния сектор промениха общоприетите разбирания за проектиране и изграждане на съвременните компютърни мрежи. В днешно време услугите и приложенията, които се разработват, изискват гъвкавост и ефективност при използване капацитета на комуникационните канали. Съществуващите мрежови архитектури имат възможността да изпълнят това условие, но с цената на много ресурси. Именно това насочи вниманието на много специалисти към набиращата все по-голяма популярност архитектурна концепция SDN, която предлага по-добро управление на мрежовата инфраструктура на бизнес организациите. В статията са представени характеристиките на SDN мрежите и основният протокол, чрез който те се реализират – OpenFlow. Специално внимание е отделено на приложните аспекти и тенденциите в развитието на SDN технологиите. Increasing information needs in the corporate sector changed the generally accepted conceptions for design and construction of modern computer networks. Nowadays, services and applications being developed require flexibility and efficiency when using the capacity of the communication channels. The existing network architectures have the potential to fulfill this requirement, but at the cost of a lot of resources. That drew the attention of many specialists to a new architectural concept – SDN, which provides a better management for the network infrastructure of business organizations. The article presents the main characteristics of the OpenFlow-based SDN networks. Focus is given to the practical aspects and trends in the SDN technologies.

“© © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. http://ieeexplore.ieee.org/document/8013750/” Energy consumption is a key concern in the deployment and operation of current data networks, for which Software-Defined Networks (SDN) have become a promising alternative. Although several works have been proposed to improve the energy efficiency, these techniques may lead to performance degradations when QoS requirements are neglected. Inspired by this problem, this letter introduces a new routing strategy, jointly considering QoS requirements and energy awareness in SDN with in-band control traffic. To that end, we present a complete formulation of the optimization problem and implement a Multi- Objective Evolutionary Algorithm. Simulation results validate the performance improvement on critical network parameters.

The 5th generation of mobile networks (5G) will enable access to information anywhere and anytime to anyone and anything, i.e., the so-called Networked Society. The details of 5G are still the subject of ongoing research and debate, mostly focused on understanding the radio technologies that can enable the 5G vision. On the other hand, less work has been dedicated so far to address the challenges that 5G will pose to the transport network infrastructure. These challenges include: (i) the capability of supporting very high aggregated traffic volumes, (ii) ensuring extreme Software-Defined Networks (SDN) aim at decoupling the control and data planes, offering programmability, scalability and resiliency to the network. For this reason, this architecture is very attractive for the 5G networks, since the services and resources are provided in real time by a centralized controller. 5G networks will cover the new advances of technology, and the end-to-end latency is one of the main constrains in this context. Hence, it is important to consider the link delay in several applications such as resource provisioning or network monitoring. In this thesis an overview of the Open Networks Operating System (ONOS), a distributed multi-layer OpenFlow controller, is provided. Furthermore, a solution for real-time link delay evaluation has been developed and tested in different network scenarios. El paradigma Software-Defined Networks (SDN) se basa en desacoplar los planos de control y datos, ofreciendo programabilidad, escalabilidad y resistencia a la red. Es por eso que esta arquitectura es muy atractiva para las redes que soporten tecnología 5G, ya que los servicios y recursos se pueden proporcionar a tiempo real utilizando un controlador central. Las redes 5G soportaran los avances tecnológicos, y la latencia es una restricción exigente en este contexto. Como consecuencia, es importante considerar este parámetro en diversas aplicaciones como el despliegue de recursos o la monitorización de redes. En esta tesis, en primer lugar, se realiza una descripción del Open Networks Operating System (ONOS), un controlador OpenFlow distribuido. Además, una solución para evaluar la latencia de los enlaces de la red a tiempo real se ha desarrollado y probado utilizando diferentes entornos de evaluación. El paradigma Software-Defined Networks (SDN) es basa en desacoblar els plans de control i de dades, oferint programabilitat, escalabilitat i resistència a les xarxes. És per això que aquesta arquitectura és molt atractiva per les xarxes 5G, ja que els serveis i recursos es poden proporcionar a temps real utilitzant un controlador central. Les xarxes 5G suportaran els avenços tecnològics, i la latència és una restricció exigent en aquest context. Com a conseqüència, és important considerar aquest paràmetre en diverses aplicacions com el desplegament de recursos o monitorització de xarxes. En aquesta tesi s'ofereix una descripció de l'Open Networks Operating System (ONOS), un controlador OpenFlow distribuït. A més, una solució per avaluar la latència dels enllaços de la xarxa a temps real s'ha desenvolupat i provat utilitzant diferents entorns.