The Experts below are selected from a list of 14955 Experts worldwide ranked by ideXlab platform

Sanner Jean-michel - One of the best experts on this subject based on the ideXlab platform.

  • Architecture du plan de contrôle SDN et placement de services réseaux dans les infrastructures des opérateurs
    HAL CCSD, 2019
    Co-Authors: Sanner Jean-michel
    Abstract:

    The evolution of telecommunications operators’ infrastructures towards the SDN and NFVparadigms requires to surmount various technical barriers. On one hand, it is necessary to deal withthe centralization of control functions, and on the other hand with the constraints of approaches coming directly from Cloud Computing. In this thesis, we addressed two issues. Firstly, we tried to define a SDN architecture more suited to the requirement of operators. For this purpose, we proposed a distributed and flexible SDN control plane to overcome the limitations of the centralized OpenFlow protocol, as well as the constraints of Network function virtualization. The proposed architecture allows for the differentiated composition, validation and deployment of dynamically Reconfigurable Network services, taking into account the SLAs associated with the services. We have illustrated some of its characteristics, namely, distribution, composition, dynamicity in a proof of concept. Secondly, to achieve the expected SLAs, we try to optimize the placement of Network services in this infrastructure. We first dealt with the issue of SDN controllers placement seeking for the optimization of latency, load and reliability metrics. Then, we considered the placement of virtualized Network functions chains. We have therefore demonstrated the potentialities and performances of evolutionary algorithms with the perspective to propose a generic resolution tool for placement of Network functions.Le contexte de l’évolution des infrastructures des opérateurs de télécommunications vers les paradigmes SDN et NFV nécessite de lever différents verrous techniques, liés d’une part à la centralisation des fonctions de contrôle, d’autre part aux contraintes d’approches qui s’inspirent directement du Cloud Computing. Dans cette thèse, nous avons abordés deux problématiques. Dans la première nous cherchons à définir une architecture SDN plus adaptée et performante par rapport aux besoins des opérateurs. Pour cela, nous avons proposé un plan de contrôle SDN distribué et flexible visant à dépasser les limites du protocole OpenFlow centralisé ainsi que les contraintes de la virtualisation des fonctions réseaux. L’architecture proposée permet la composition, puis la validation et le déploiement différenciés de services réseaux composables et Reconfigurables dynamiquement en prenant en compte les SLA associés aux services. Nous avons illustré certaines propriétés de cette architecture, distribution, composition, dynamicité dans une preuve de concepts. Dans la deuxième, pour réaliser les SLA attendus, nous cherchons à optimiser le placement des services réseaux dans cette infrastructure. Nous avons d’abord traité la problématique du placement de contrôleurs SDN en optimisant des métriques de latence, de charge et de fiabilité, puis de manière plus générale le placement de chaînes de fonctions réseaux virtualisées. Nous avons démontré pour cela les potentialités et les performances des algorithmes évolutionnaires pour tenter de proposer un outil de résolution générique de placement de fonctions réseaux

  • Architecture du plan de contrôle SDN et placement des services réseaux dans les infrastructures des opérateurs
    HAL CCSD, 2019
    Co-Authors: Sanner Jean-michel
    Abstract:

    The evolution of telecommunications operators’ infrastructures towards the SDN and NFV paradigms requires to surmount various technical barriers. On one hand, it is necessary to deal with the centralization of control functions, and on the other hand with the constraints of approaches coming directly from Cloud Computing. In this thesis, we addressed two issues. Firstly, we tried to define a SDN architecture more suited to the requirement of operators. For this purpose, we proposed a distributed and flexible SDN control plane to overcome the limitations of the centralized OpenFlow protocol, as well as the constraints of Network function virtualization. The proposed architecture allows for the differentiated composition, validation and deployment of dynamically Reconfigurable Network services, taking into account the SLAs associated with the services. We have illustrated some of its characteristics, namely, distribution, composition, dynamicity in a proof of concept. Secondly, to achieve the expected SLAs, we try to optimize the placement of Network services in this infrastructure. We first dealt with the issue of SDN controllers placement seeking for the optimization of latency, load and reliability metrics. Then, we considered the placement of virtualized Network functions chains. We have therefore demonstrated the potentialities and performances of evolutionary algorithms with the perspective to propose a generic resolution tool for placement of Network functions.Le contexte de l'évolution des infrastructures des opérateurs de télécommunications vers les paradigmes SDN et NFV nécessite de lever différents verrous techniques, liés d'une part à la centralisation des fonctions de contrôle, d'autre part aux contraintes d'approches qui s'inspirent directement du Cloud Computing. Dans cette thèse, nous avons abordés deux problématiques. Dans la première nous cherchons à définir une architecture SDN plus adaptée et performante par rapport aux besoins des opérateurs. Pour cela, nous avons proposé un plan de contrôle SDN distribué et flexible visant à dépasser les limites du protocole OpenFlow centralisé ainsi que les contraintes de la virtualisation des fonctions réseaux. L'architecture proposée permet la composition, puis la validation et le déploiement différenciés de services réseaux composables et Reconfigurables dynamiquement en prenant en compte les SLA associés aux services. Nous avons illustré certaines propriétés de cette architecture, distribution, composition, dynamicité dans une preuve de concepts. Dans la deuxième, pour réaliser les SLA attendus, nous cherchons à optimiser le placement des services réseaux dans cette infrastructure. Nous avons d'abord traité la problématique du placement de contrôleurs SDN en optimisant des métriques de latence, de charge et de fiabilité, puis de manière plus générale le placement de chaînes de fonctions réseaux virtualisées. Nous avons démontré pour cela les potentialités et les performances des algorithmes évolutionnaires pour tenter de proposer un outil de résolution générique de placement de fonctions réseaux

  • SDN control plane architecture and placement of Network services in TelCo infrastructures
    2019
    Co-Authors: Sanner Jean-michel
    Abstract:

    Le contexte de l'évolution des infrastructures des opérateurs de télécommunications vers les paradigmes SDN et NFV nécessite de lever différents verrous techniques, liés d'une part à la centralisation des fonctions de contrôle, d'autre part aux contraintes d'approches qui s'inspirent directement du Cloud Computing. Dans cette thèse, nous avons abordés deux problématiques. Dans la première nous cherchons à définir une architecture SDN plus adaptée et performante par rapport aux besoins des opérateurs. Pour cela, nous avons proposé un plan de contrôle SDN distribué et flexible visant à dépasser les limites du protocole OpenFlow centralisé ainsi que les contraintes de la virtualisation des fonctions réseaux. L'architecture proposée permet la composition, puis la validation et le déploiement différenciés de services réseaux composables et Reconfigurables dynamiquement en prenant en compte les SLA associés aux services. Nous avons illustré certaines propriétés de cette architecture, distribution, composition, dynamicité dans une preuve de concepts. Dans la deuxième, pour réaliser les SLA attendus, nous cherchons à optimiser le placement des services réseaux dans cette infrastructure. Nous avons d'abord traité la problématique du placement de contrôleurs SDN en optimisant des métriques de latence, de charge et de fiabilité, puis de manière plus générale le placement de chaînes de fonctions réseaux virtualisées. Nous avons démontré pour cela les potentialités et les performances des algorithmes évolutionnaires pour tenter de proposer un outil de résolution générique de placement de fonctions réseaux.The evolution of telecommunications operators’ infrastructures towards the SDN and NFV paradigms requires to surmount various technical barriers. On one hand, it is necessary to deal with the centralization of control functions, and on the other hand with the constraints of approaches coming directly from Cloud Computing. In this thesis, we addressed two issues. Firstly, we tried to define a SDN architecture more suited to the requirement of operators. For this purpose, we proposed a distributed and flexible SDN control plane to overcome the limitations of the centralized OpenFlow protocol, as well as the constraints of Network function virtualization. The proposed architecture allows for the differentiated composition, validation and deployment of dynamically Reconfigurable Network services, taking into account the SLAs associated with the services. We have illustrated some of its characteristics, namely, distribution, composition, dynamicity in a proof of concept. Secondly, to achieve the expected SLAs, we try to optimize the placement of Network services in this infrastructure. We first dealt with the issue of SDN controllers placement seeking for the optimization of latency, load and reliability metrics. Then, we considered the placement of virtualized Network functions chains. We have therefore demonstrated the potentialities and performances of evolutionary algorithms with the perspective to propose a generic resolution tool for placement of Network functions

E Modiano - One of the best experts on this subject based on the ideXlab platform.

  • quantifying the benefit of configurability in circuit switched wdm ring Networks with limited ports per node
    Journal of Lightwave Technology, 2001
    Co-Authors: B Schein, E Modiano
    Abstract:

    In a Reconfigurable Network, lightpath connections can be dynamically changed to reflect changes in traffic conditions. This paper characterizes the gain in traffic capacity that a Reconfigurable wavelength division multiplexed (WDM) Network offers over a fixed topology Network where lightpath connections are fixed and cannot be changed. We define the gain as the ratio of the maximum offered loads that the two systems can support for a given blocking probability. We develop a system model to approximate the blocking probability for both the fixed and Reconfigurable systems. This model is different from previous models developed to analyze the blocking probability in WDM Networks in that it accounts for a port limitation at the nodes. We validate our model via simulation and find that it agrees strongly with simulation results. We study high-bandwidth calls, where each call requires an entire wavelength and find that reconfigurability offers a substantial performance improvement, particularly when the number of available wavelengths significantly exceeds the number of ports per node. In this case, in a ring with N nodes, the gain approaches a factor of N/2 over a fixed topology unidirectional ring, and N/4 over a fixed topology bidirectional ring. Hence, a Reconfigurable unidirectional (bidirectional) ring can support N/2(N/4) times the load of a fixed topology unidirectional (bidirectional) ring. We also show that for a given traffic load, a configurable system requires far fewer ports per node than a fixed topology system. These port savings can potentially result in a significant reduction in overall system costs.

  • quantifying the benefit of configurability in circuit switched wdm ring Networks
    International Conference on Computer Communications, 2000
    Co-Authors: B Schein, E Modiano
    Abstract:

    We attempt to characterize the gain in traffic capacity that a Reconfigurable Network offers over a fixed topology Network. We define the gain as the ratio of the maximum offered loads that the two systems can support for a given blocking probability. We develop a system model to analytically predict the blocking probability for both the fixed and Reconfigurable systems. This model is different from previous models developed to analyze the blocking probability in WDM Networks in that it accounts for a port limitation at the nodes. We study high-bandwidth calls, where each call requires an entire wavelength. We find that reconfigurability offers a substantial performance improvement, particularly when the number of available wavelengths significantly exceeds the number of ports per node. In this case, we find that the gain approaches a factor of N/2 over a fixed topology unidirectional ring, and N/4 over a fixed topology bi-directional ring (where N is the number of nodes in the ring). We validate our model via simulation, and we find that it agrees strongly with the simulation results, particularly for a large number of ports per node. We also obtain upper and lower bounds on capacity for various ring topologies that give additional insight into the benefits of configurability.

B Schein - One of the best experts on this subject based on the ideXlab platform.

  • quantifying the benefit of configurability in circuit switched wdm ring Networks with limited ports per node
    Journal of Lightwave Technology, 2001
    Co-Authors: B Schein, E Modiano
    Abstract:

    In a Reconfigurable Network, lightpath connections can be dynamically changed to reflect changes in traffic conditions. This paper characterizes the gain in traffic capacity that a Reconfigurable wavelength division multiplexed (WDM) Network offers over a fixed topology Network where lightpath connections are fixed and cannot be changed. We define the gain as the ratio of the maximum offered loads that the two systems can support for a given blocking probability. We develop a system model to approximate the blocking probability for both the fixed and Reconfigurable systems. This model is different from previous models developed to analyze the blocking probability in WDM Networks in that it accounts for a port limitation at the nodes. We validate our model via simulation and find that it agrees strongly with simulation results. We study high-bandwidth calls, where each call requires an entire wavelength and find that reconfigurability offers a substantial performance improvement, particularly when the number of available wavelengths significantly exceeds the number of ports per node. In this case, in a ring with N nodes, the gain approaches a factor of N/2 over a fixed topology unidirectional ring, and N/4 over a fixed topology bidirectional ring. Hence, a Reconfigurable unidirectional (bidirectional) ring can support N/2(N/4) times the load of a fixed topology unidirectional (bidirectional) ring. We also show that for a given traffic load, a configurable system requires far fewer ports per node than a fixed topology system. These port savings can potentially result in a significant reduction in overall system costs.

  • quantifying the benefit of configurability in circuit switched wdm ring Networks
    International Conference on Computer Communications, 2000
    Co-Authors: B Schein, E Modiano
    Abstract:

    We attempt to characterize the gain in traffic capacity that a Reconfigurable Network offers over a fixed topology Network. We define the gain as the ratio of the maximum offered loads that the two systems can support for a given blocking probability. We develop a system model to analytically predict the blocking probability for both the fixed and Reconfigurable systems. This model is different from previous models developed to analyze the blocking probability in WDM Networks in that it accounts for a port limitation at the nodes. We study high-bandwidth calls, where each call requires an entire wavelength. We find that reconfigurability offers a substantial performance improvement, particularly when the number of available wavelengths significantly exceeds the number of ports per node. In this case, we find that the gain approaches a factor of N/2 over a fixed topology unidirectional ring, and N/4 over a fixed topology bi-directional ring (where N is the number of nodes in the ring). We validate our model via simulation, and we find that it agrees strongly with the simulation results, particularly for a large number of ports per node. We also obtain upper and lower bounds on capacity for various ring topologies that give additional insight into the benefits of configurability.

Sajal K Das - One of the best experts on this subject based on the ideXlab platform.

  • peternet an emergent technology based radio access Network architecture for next generation cellular wireless systems
    Broadband Communications Networks and Systems, 2004
    Co-Authors: Samik Ghosh, Kalyan Basu, Sajal K Das
    Abstract:

    For next generation wireless cellular Networks, the radio access Network (RAN) component is destined to evolve into a high-capacity, flexible and Reconfigurable Network architecture supporting smaller cell sites and newer technologies. Based on the well-known Petersen graph, in this paper we propose a RAN architecture, named PeterNet, that interconnects the Network elements in a mesh topology. Focusing on optical wireless as the underlying transmission technology, we analyze our architecture on two key indices - carrier-class Network reliability (99.999%) and infrastructure cost - providing comparative performance analysis with existing tree/star and regular grid Networks. Furthermore, we generalize our architecture to what we call, the GPeterNet, which allows dynamic reconfiguration and redesign of the access Network while leveraging the underlying salient features.

Hans P. Graf - One of the best experts on this subject based on the ideXlab platform.

  • A Reconfigurable VLSI Neural Network
    IEEE Journal of Solid-State Circuits, 1992
    Co-Authors: Srinivas Satyanarayana, Yannis P. Tsividis, Hans P. Graf
    Abstract:

    Due to the variety of architectures that need be considered while attempting solutions to various problems using neural Networks, the implementation of a neural Network with programmable topology and programmable weights has been undertaken. A new circuit block, the distributed neuron-synapse, has been used to implement a 1024 synapse Reconfigurable Network on a VLSI chip. In order to evaluate the performance of the VLSI chip, a complete test setup consisting of hardware for configuring the chip, programming the synaptic weights, presenting analog input vectors to the chip, and recording the outputs of the chip, has been built. Following the performance verification of each circuit block on the chip, various sample problems were solved. In each of the problems the synaptic weights were determined by training the neural Network using a gradient-based learning algorithm which is incorporated in the experimental test setup. The results of this work indicate that Reconfigurable neural Networks built using distributed neuron synapses can be used to solve various problems efficiently