Le SDN (Software-Defined Networking) émerge comme une nouvelle architecture pour la programmation des réseaux. A l'origine, l'idée du SDN est de déplacer le plan de contrôle à l'extérieur des équipements, et de permettre ainsi un contrôle déporté de l'ensemble depuis une entité logicielle logique nommée "contrôleur". Le principal avantage d'une telle approche est de centraliser donc toute l'intelligence de gestion du réseau dans le contrôleur, qui s'appuie pour cela sur des protocoles standard et assure par ce biais la reprogrammation de la totalité de la partie du réseau sous son contrôle. L'évolution technologique vers le SDN est toujours en cours dans des scénarios de déploiement programmable et flexible des réseaux mobiles. Le NFV (Network Function Virtualization) est le processus de déplacement ou de migration des fonctions réseau d'un équipement dédié de réseau vers des serveurs génériques dans le Cloud. Les SDN et NFV sont deux technologies étroitement liées qui sont souvent utilisées ensemble. Le couplage fort entre les plans de contrôle et de données, ainsi que les limitations en matière de passage à l'échelle et de flexibilité, font que la virtualisation des réseaux mobiles actuels nécessite non seulement l'utilisation du Cloud Computing mais aussi les récentes innovations telles que SDN et NFV pour pouvoir permettre un déploiement à la demande des services réseaux (Network-as-a-Service) aux utilisateurs. Les lignes de recherche globales de cette thèse s'inscrivent dans deux principaux cas d'utilisation. Ces cas d'utilisation, bien qu'appelés de la "prochaine génération de réseaux mobiles", sont le "Telco" et le "Vertical", qui apparaissent ici couplés, les deux étant traditionnellement complètement séparés. Dans les cas d'utilisation de "télécommunications", nous exploitons les avantages de SDN pour avoir un cadre de contrôle flexible pour les réseaux d'auto-organisation (SON) et la division de traitement dynamique des utilisateurs. Dans le cas d'utilisation de "verticale", nous appliquons divers avantages du protocole SDN et OpenFlow pour utiliser efficacement les ressources radio du réseau de backhaul dans le système de communication train-sol. Notre cadre d'étude du SDN, en général, peut être une solution efficace et alternative pour la gestion RAN (Radio Access Network), c'est-à-dire pour des objectives comme l'optimisation des ressources radio, l'optimisation du réseau, la gestion de la mobilité et l'équilibrage de la charge, peuvent être atteint avec ce cadre. Grâce à l'analyse et l'expérimentation concrète des SDN et NFV pour le RAN, nous montrons que les solutions proposées dans ce travail peuvent apporter un faisceau d'avantages évidents aux réseaux mobiles tels que la flexibilité, la programmabilité, la gestion unifiée et la mise en œuvre de nouveaux services / Software-Defined Networking (SDN) has emerged as a new intelligent architecture for network programmability. The primary idea behind SDN is to move the control-plane outside the switches and enable external control of data-plane through a logical software entity called controller. Such approach benefits mobile network management by brining complete intelligence to the logically centralized controller. Network Function Virtualization (NFV) is the process of relocating or migrating network functions from dedicated hardware to generic servers. SDN and NFV are two closely related technologies that are often used together. The traditional mobile network architecture due to its strongest coupling between control and data planes along with limitations in scalability and flexibility requires the usage of cloud computing along with the recent revolutionary approaches in networking such as SDN and NFV to have an architecture that deploys on demand "Network-as-a-Service" for users. The global research focus of this thesis falls in to two main use cases of next generation mobile networks such as Telco and Vertical. In the telco use cases, we exploit the advantages of SDN to have flexible control framework for both Self-Organizing Networks (SON) and dynamic user processing split. In vertical use case, we apply various advantages of SDN and OpenFlow protocol to efficiently utilize the scare radio resources of wireless backhaul network in the train-to-ground communication system. Our SDN framework in general can be an efficient and alternative solution for RAN management i.e. Radio Optimization, Network Optimization, Mobility Management and Load Balancing can be achieved with such framework. Through analysis and experimentation of SDN frameworks for RAN, we shows that the proposed solutions can bring set of advantages to wireless networks such as flexibility, programmability, unified management, and enables new services

Dans cette thèse, on présente nos travaux sur la virtualisation dans le contexte de la réplication des serveurs de contenu vidéo. Les travaux couvrent la conception d’une architecture de virtualisation pour laquelle on présente aussi plusieurs algorithmes qui peuvent réduire les couts globaux à long terme et améliorer la performance du système. Le travail est divisé en plusieurs parties : les solutions optimales, les solutions heuristiques pour des raisons de passage à l’échelle, l’orchestration des services, l’optimisation multi-objective, la planification de services dans des réseaux actifs et complexes et l'intégration d'algorithmes dans une plate-forme réelle / In this thesis, we designed and implemented a tool which performs optimizations that reduce the number of migrations necessary for a delivery task. We present our work on virtualization in the context of replication of video content servers. The work covers the design of a virtualization architecture for which there are also several algorithms that can reduce overall costs and improve system performance. The thesis is divided into several parts: optimal solutions, greedy (heuristic) solutions for reasons of scalability, orchestration of services, multi-objective optimization, service planning in complex active networks, and integration of algorithms in real platform. This thesis is supported by models, implementations and simulations which provide results that showcase our work, quantify the importance of evaluating optimization techniques and analyze the trade-off between reducing operator cost and enhancing end user satisfaction index

Network functions (NFs) hold a key role in networks, offering in-network services, such as enhanced performance, policy enforcement, and security. Traditionally, NFs have been implemented in specialized, thus expensive hardware. To lower the costs of deploying NFs, network operators have adopted network functions virtualization (NFV), by migrating NFs from hardware to software running in commodity servers. Several approaches to NFV have shown that commodity network stacks and drivers (e.g., Linux-based) struggle to keep up with increasing hardware speed. Despite this, popular networking services still rely on these commodity components. Moreover, chaining NFs (also known as service chaining) is challenging due to redundancy in the elements of the chain. This licentiate thesis addresses the performance problems of NFV service chains.The first contribution is a framework that (i) profiles NFV service chains to uncover performance degradation reasons and (ii) leverages the profiler’s data to accelerate these chains, by combining multiplexing of system calls with scheduling strategies. These accelerations improve the cache utilization and thereby the end-to-end latency of chained NFs is reduced by a factor of three. Moreover, the same chains experience a multi-fold latency variance reduction; this result improves the quality of highly-interactive services.The second contribution of this thesis substantially revises the way NFV service chains are realized. NFV service chains are synthesized while eliminating redundant input/output and repeated elements, providing consolidated stateful cross layer packet operations across the chain. This software-based synthesis achieves line-rate 40 Gbps throughput for stateful and long service chains. This performance is 8.5x higher than the performance achieved by the software-based state of the art FastClick framework. Experiments with three example Internet Service Provider-level service chains show that this synthesis approach operates at 40 Gbps, when the classification of these chains is offloaded to an OpenFlow switch. / Nätverksfunktioner (NF) har en nyckelroll i nätverk. De erbjuder tjänster i nätverken som förbättrad prestanda, policy övervakning och säkerhetsfunktioner. Vanligtvis så har NF implementerats med hjälp av specialiserad, och därmed kostsam, hårdvara. Detta har lett till att nätverksoperatörer har börjat använda nätverksfunktionsvirtualisering (NFV) för att minska kostnaden. NFV implementeras genom att NF flyttas från specialiserad hårdvara till mjukvara som kör på vanliga servrar. Flera försök med NFV har visat att vanliga nätverksstackar och drivrutiner (exempelvis Linux baserade) har svårt att erbjuda samma prestanda som hårdvaran gör. Trots detta bygger flera populära nätverkstjänster på NFV. Dessutom är det en utmaning att koppla samman NFV i kedjor, då redundanta operationer utförs. I den här avhandlingen försöker vi lösa prestanda problem kopplade till kedjor av NFV. Det första bidraget i den här avhandlingen är ett ramverk som (i) profilerar NFV kedjor för att hitta orsaker till prestanda problem samt (ii) använder profileringsdata för att förbättra prestandan i kedjorna. Detta görs genom att kombinera multiplexing av systemanrop med planläggningsstrategier. Tillsammans förbättrar dessa lösningar cache användningen och minskar därmed end-to-end latensen i kedjade NFV med en faktor tre. Dessutom minskar vår metod variansen i latens, något som är viktigt för tjänstekvalitén i interaktiva tjänster.Det andra bidraget i den här avhandlingen är en omarbetning av hur kedjade NFV konstrueras. Vi syntetiserar NFV service kedjor genom att ta bort redundanta element och konsoliderar paketoperationer som sträcker sig över flera lager i nätverksstacken. Vår mjukvarubaserade lösning klarar av 40 Gbps genomströmning i en lång kedja. Detta är 8.5 ggr mer än vad som uppnåtts med den tidigare standard lösningen för mjukvara, ramverket FastClick. Vi presenterar experiment med tre servicekedjor för nätverksleverantörer där vår syntetiserade lösning klarar 40 Gbps, när klassificeringen av kedjan görs med hjälp av en OpenFlow switch. / <p>QC 20161103</p> / European Union Horizon 2020 BEhavioural BAsed forwarding (BEBA) / European Research Council (ERC) PROPHET

The telecom sector contributes to 1.6% of global emissions and is expanding to meet the increasing demand, raising concerns regarding energy consumption and CO2 emissions. This study investigates the potential of Network Function Virtualization (NFV) technology to decrease the operating carbon footprint and additionally aid to reach NetZero goals by 2040. A comparative case study with a network located in an African country evaluated the Mobile Switching Center Server (MSC-S) in a Physical Network Function (PNF) versus NFV deployment. To evaluate the results a Green ICT evaluation framework has been utilized, containing three sections Green ICT Strategy, Green ICT Practice, and Green ICT Measure, and has been adjusted to fit the context of the case. Initial results showed that the MSC-S with the NFV deployment consumes significantly more energy than the previous PNF solution. This was suspected to be a consequence of non-optimized hardware utilization. In a scenario analysis, it was shown that increased hardware efficiency for the NFV technology leads to lower operating energy consumption. However, the derived configurations could potentially lead to redundancy and service issues. The results from both the base case and scenario analysis were further evaluated with the refined framework. A final framework was suggested considering Green ICT Learnings and Contextual factors to facilitate the evaluation and strategy for future Green ICT initiatives. / Telekomsektorn bidrar till 1,6% av de globala koldioxidutsläppen och expanderar för att möta den ökande efterfrågan, vilket bidragit till en större oro kring energiförbrukning och CO2-utsläpp. Denna studie undersöker potentialen hos NFV teknik för att minska det operativa koldioxidavtrycket samt bidra till att uppnå NetZero-målen 2040. I en jämförande fallstudie av ett telekomnätverk i ett afrikanskt land utvärderades en MSC-S applikation i en PNF kontra NFV installation. För att utvärdera resultaten har ett utvärderingsramverk för Grön ICT använts, som innehåller tre delar: Grön ICT-strategi, Grön ICT-praktik och Grön ICT-mått. Vidare har ramverket anpassats för att kunna appliceras på fallstudien. Resultaten från den initiala basfallsstudien visade att MSC-S med NFV-utbyggnaden förbrukar betydligt mer energi än den tidigare PNF-lösningen. Detta misstänktes vara en konsekvens av en icke-optimerad hårdvaruanvändning. I en scenarioanalys visades att ökad hårdvarueffektivitet för NFV-tekniken leder till lägre driftsenergiförbrukning. De föreslagna konfigurationerna, med högre hårdvarueffektivitet, bidrar visserligen till potentiella redundans- och serviceproblem. Resultaten från både basfalls- och scenarioanalysen utvärderades ytterligare med det förfinade ramverket för Grön ICT. Ett slutligt ramverk föreslogs som tog hänsyn till Gröna ICT-lärdommar och kontextuella faktorer. Syftet med det slutgiltiga ramverket är att underlätta utvärdering av strategier för framtida Gröna ICT-initiativ

Growth in the Information and Communication Technology sector is increasing the need to improve the quality of service and energy efficiency, as this industry has already surpassed 12% of global energy consumption in 2017. Data centers correspond to a large part of this consumption, accounting for about 15% of energy expenditure on the Information and Communication Technology domain; moreover, the subsystem that generates the most costs for data center operators is that of servers and storage. Many solutions have been proposed to reduce server consumption, such as the use of dynamic voltage and frequency scaling, a technology that enables the adaptation of energy consumption to the workload by modifying the operating voltage and frequency, although they are not optimized for network traffic. In this thesis, a control method was developed using a prediction engine based on the analysis of the ongoing traffic. Machine learning algorithms based on Neural Networks and Support Vector Machines have been used, and it was verified that it is possible to reduce power consumption by up to 12% on servers with Intel Sandy Bridge processor and up to 21 % in servers with Intel Haswell processor when compared to the maximum frequency, which is currently the most used solution in the industry. / O crescimento do setor de Tecnologia da Informação e Comunicação está aumentando a necessidade de melhorar a qualidade de serviço e a eficiência energética, pois o setor já ultrapassou a marca de 12% do consumo energético global em 2017. Data centers correspondem a grande parte desse consumo, representando cerca de 15% dos gastos com energia do setor Tecnologia Informação e Comunicação; além disso, o subsistema que gera mais custos para operadores de data centers é o de servidores e armazenamento. Muitas soluções foram propostas a fim de reduzir o consumo de energia com servidores, como o uso de escalonamento dinâmico de tensão e frequência, uma tecnologia que permite adaptar o consumo de energia à carga de trabalho, embora atualmente não sejam otimizadas para o processamento do tráfego de rede. Nessa dissertação, foi desenvolvido um método de controle usando um mecanismo de previsão baseado na análise do tráfego que chega aos servidores. Os algoritmos de aprendizado de máquina baseados em Redes Neurais e em Máquinas de Vetores de Suporte foram utilizados, e foi verificado que é possível reduzir o consumo de energia em até 12% em servidores com processador Intel Sandy Bridge e em até 21% em servidores com processador Intel Haswell quando comparado com a frequência máxima, que é atualmente a solução mais utilizada na indústria.

Growth in the Information and Communication Technology sector is increasing the need to improve the quality of service and energy efficiency, as this industry has already surpassed 12% of global energy consumption in 2017. Data centers correspond to a large part of this consumption, accounting for about 15% of energy expenditure on the Information and Communication Technology domain; moreover, the subsystem that generates the most costs for data center operators is that of servers and storage. Many solutions have been proposed to reduce server consumption, such as the use of dynamic voltage and frequency scaling, a technology that enables the adaptation of energy consumption to the workload by modifying the operating voltage and frequency, although they are not optimized for network traffic. In this thesis, a control method was developed using a prediction engine based on the analysis of the ongoing traffic. Machine learning algorithms based on Neural Networks and Support Vector Machines have been used, and it was verified that it is possible to reduce power consumption by up to 12% on servers with Intel Sandy Bridge processor and up to 21 % in servers with Intel Haswell processor when compared to the maximum frequency, which is currently the most used solution in the industry. / O crescimento do setor de Tecnologia da Informação e Comunicação está aumentando a necessidade de melhorar a qualidade de serviço e a eficiência energética, pois o setor já ultrapassou a marca de 12% do consumo energético global em 2017. Data centers correspondem a grande parte desse consumo, representando cerca de 15% dos gastos com energia do setor Tecnologia Informação e Comunicação; além disso, o subsistema que gera mais custos para operadores de data centers é o de servidores e armazenamento. Muitas soluções foram propostas a fim de reduzir o consumo de energia com servidores, como o uso de escalonamento dinâmico de tensão e frequência, uma tecnologia que permite adaptar o consumo de energia à carga de trabalho, embora atualmente não sejam otimizadas para o processamento do tráfego de rede. Nessa dissertação, foi desenvolvido um método de controle usando um mecanismo de previsão baseado na análise do tráfego que chega aos servidores. Os algoritmos de aprendizado de máquina baseados em Redes Neurais e em Máquinas de Vetores de Suporte foram utilizados, e foi verificado que é possível reduzir o consumo de energia em até 12% em servidores com processador Intel Sandy Bridge e em até 21% em servidores com processador Intel Haswell quando comparado com a frequência máxima, que é atualmente a solução mais utilizada na indústria.

Telecom service providers are adopting a Network Functions Virtualization (NFV) based service delivery model, in response to the unprecedented traffic growth and an increasing customers demand for new high-quality network services. In NFV, telecom network functions are virtualized and run on top of commodity servers. Ensuring network performance equivalent to the legacy non-virtualized system is a determining factor for the success of telecom networks virtualization. Whereas in virtualized systems, achieving carrier-grade network performance such as low latency, high throughput, and high availability to guarantee the quality of experience (QoE) for customer is challenging. In this thesis, we focus on addressing the latency challenge. We investigate the delay overhead of virtualization by comprehensive network performance measurements and analysis in a controlled virtualized environment. With this, a break-down of the latency incurred by the virtualization and the impact of co-locating virtual machines (VMs) of different workloads on the end-to-end latency is provided. We exploit this result to develop an optimization model for placement and provisioning of the virtualized telecom network functions to ensure both the latency and cost-efficiency requirements. To further alleviate the latency challenge, we propose a multipath transport protocol MDTCP, that leverage Explicit Congestion Notification (ECN) to quickly detect and react to an incipient congestion to minimize queuing delays, and achieve high network utilization in telecom datacenters. / HITS, 4707

With the upcoming 5G deployment and the exponentially increasing data transmitted over cellular networks, off the shelf hardware won't provide enough performance to cope with the data being transferred over cellular networks. To tackle that problem, hardware accelerators will be of great support thanks to their better performances and lower energy consumption. However, hardware accelerators are not a silver bullet as their very nature prevents them to be as flexible as CPUs. Hardware accelerators integration into Kubernetes and Docker, respectively the most used tools for orchestration and containerization, is still not as flexible as it would need. In this thesis, we developed a framework that allows for a more flexible integration of these accelerators into a Kubernetes cluster using Docker containers making use of an abstraction layer instead of the classic virtualization process. Our results compare the performance of an execution with and without the framework that was developed during this thesis. We found that the framework's overhead depends on the size of the data being processed by the accelerator but does not go over a very low percentage of the total execution time. This framework provides an abstraction for hardware accelerators and thus provides an easy way to integrate hardware accelerated applications into a heterogeneous cluster or even across different clusters with different hardware accelerators types. This framework also moves the hardware specific parts of an accelerated program from the containers to the infrastructure and enables a new kind of service, OpenCL as a service.

One recent innovation in the networking industry, is the concept of Network FunctionVirtualization (NFV). NFV is based on a networking paradigm in which network functions,which have typically been implemented in the form of dedicated hardware appliances in thepast, are implemented in software and deployed on commodity hardware using modernvirtualization techniques. While the most common approach is to place each virtual networkfunction in a virtual machine - using hardware-level virtualization – the growing influenceand popularity of Docker and other container-based solutions has naturally led to the idea ofcontainerized deployments. This is a promising concept, as containers (or operating systemlevel virtualization) can offer a flexible and lightweight alternative to hardware-levelvirtualization, with the ability to use the resources of the host directly. The main problem withthis concept, is the fact that the default behavior of Docker and similar technologies is to relyon the networking stack of the host, which typically isn’t performant enough to handle theperformance requirements associated with NFV. In this dissertation, an attempt is made toevaluate the feasibility of using userspace networking to accelerate the network performanceof Docker containers, bypassing the standard Linux networking stack by moving the packetprocessing into userspace.

Next generation mobile networks are designed to run in a virtualized environment, enabling rapid infrastructure deployment and high flexibility for coping with increasing traffic demands and new service requirements. Such network function virtualization imposes additional packet latencies and potential bottlenecks not present in legacy network equipment when run on dedicated hardware; such bottlenecks include PCIe transfer delays, virtualization overhead, and utilizing commodity server hardware which is not optimized for packet processing operations.Through recent developments in P4 programmable networking devices, it is possible to implement complex packet processing pipelines directly in the network data plane; allowing critical traffic flows to be offloaded and flexibly hardware accelerated on new programmable packet processing hardware, prior to entering the virtualized environment.In this thesis, we design and implement a novel hybrid NFV processing architecture which integrates programmable NICs and commodity server hardware, capable of offloading virtual network functions for specified traffic flows directly to the server network card; allowing these flows to completely bypass softwarization overhead, while less sensitive traffic process on the underlying host server.An evaluation in a testbed with customized traffic generators show that accelerated flows have significantly lower jitter and latency, compared with flows processed on commodity server hardware. Our evaluation gives important insights into the designs of such hardware accelerated virtual network deployments, showing that hybrid network architectures are a viable solution for enabling infrastructure scalability without sacrificing critical flow performance.