Emerging 5G mobile networks are envisioned to become multi-service environments, enabling the dynamic deployment of services with a diverse set of performance requirements, accommodating the needs of mobile network operators, verticals and over-the-top service providers. The Radio Access Network (RAN) part of mobile networks is expected to play a very significant role towards this evolution. Unfortunately, such a vision cannot be efficiently supported by the conventional RAN architecture, which adopts a fixed and rigid design. For the network to evolve, flexibility in the creation, management and control of the RAN components is of paramount importance. The key elements that can allow us to attain this flexibility are the programmability and the virtualization of the network functions. While in the case of the mobile core, these issues have been extensively studied due to the advent of technologies like Software-Defined Networking (SDN) and Network Functions Virtualization (NFV) and the similarities that the core shares with other wired networks like data centers, research in the domain of the RAN is still in its infancy. The contributions made in this thesis significantly advance the state of the art in the domain of RAN programmability and virtualization in three dimensions. First, we design and implement a software-defined RAN (SD-RAN) platform called FlexRAN, that provides a flexible control plane designed with support for real-time RAN control applications, flexibility to realize various degrees of coordination among RAN infrastructure entities, and programmability to adapt control over time and easier evolution to the future following SDN/NFV principles. Second, we leverage the capabilities of the FlexRAN platform to design and implement Orion, which is a novel RAN slicing system that enables the dynamic on-the-fly virtualization of base stations, the flexible customization of slices to meet their respective service needs and which can be used in an end-to-end network slicing setting. Third, we focus on the use case of multi-tenancy in a neutral-host indoors small-cell environment, where we design Iris, a system that builds on the capabilities of FlexRAN and Orion and introduces a dynamic pricing mechanism for the efficient and flexible allocation of shared spectrum to the tenants. A number of additional use cases that highlight the benefits of the developed systems are also presented. The lessons learned through this research are summarized and a discussion is made on interesting topics for future work in this domain. The prototype systems presented in this thesis have been made publicly available and are being used by various research groups worldwide in the context of 5G research.

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

Orientador: Edmundo Roberto Mauro Madeira / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-25T12:25:32Z (GMT). No. of bitstreams: 1 Rosa_RaphaelVicente_M.pdf: 3185742 bytes, checksum: 0fe239607d96962513c068e7379c6f19 (MD5) Previous issue date: 2014 / Resumo: Atualmente provedores de infraestrutura (Infrastructure Providers - InPs) alocam recursos virtualizados, computacionais e de rede, de seus data centers para provedores de serviços na forma de data centers virtuais (Virtual Data Centers - VDCs). Almejando maximizar seus lucros e usar de forma eficiente os recursos de seus data centers, InPs lidam com o problema de otimizar a alocação de múltiplos VDCs. Mesmo que a alocação de máquinas virtuais em servidores seja feita de maneira otimizada por diversas técnicas e algoritmos já existentes, aplicações de computação em nuvem ainda tem o desempenho prejudicado pelo gargalo do subaproveitamento de recursos de rede, explicitamente definidos por limitações de largura de banda e latência. Baseado no paradigma de Redes Definidas por Software, nós aplicamos o modelo de rede como serviço (Network-as-a-Service - NaaS) para construir uma arquitetura de data center bem definida para dar suporte ao problema de aprovisionamento de redes virtuais em data centers. Construímos serviços sobre o plano de controle da plataforma RouteFlow os quais tratam a alocação de redes virtuais de data center otimizando a utilização de recursos da infraestrutura de rede. O algoritmo proposto neste trabalho realiza a tarefa de alocação de redes virtuais, baseado na agregação de informações de um plano virtual executando o protocolo BGP eficientemente mapeado a uma topologia física de rede folded-Clos definida por \textit{switches} com suporte a OpenFlow 1.3. Em experimentos realizados, mostramos que o algoritmo proposto neste trabalho realiza a alocação de redes virtuais de data center de forma eficiente, otimizando o balanceamento de carga e, consequentemente, a utilização de recursos da infraestrutura de rede de data centers. A estratégia de alocação de largura de banda utilizada demonstra flexibilidade e simplicidade para atender a diferentes padrões de comunicação nas redes virtuais ao mesmo tempo que permite elasticidade ao balanceamento de carga na rede. Por fim, discutimos como a arquitetura e algoritmo propostos podem ser estendidos para atender desempenho, escalabilidade, e outros requisitos de arquiteturas de redes de data center / Abstract: Nowadays infrastructure providers (InPs) allocate virtualized resources, computational and network, of their data center to service providers (SPs) in the form of virtual data centers (VDCs). Aiming maximize revenues and thus efficiently use the resources of their virtualized data centers, InPs handle the problem to optimally allocate multiple VDCs. Even if the allocation of virtual machines in servers can be made using well known techniques and algorithms already existent, cloud computing applications still have performance limitations imposed by the bottleneck of network resources underutilization, which are explicitly defined by bandwidth and latency constraints. Based on Software Defined Network paradigm we apply the Network-as-a-Service model to build a data center network architecture well-suited to the problem of virtual networks embedding. We build services over the control plane of the RouteFlow platform that perform the allocation of virtual data center networks optimizing the utilization of network infrastructure resources. This task is performed by the algorithm proposed in this dissertation, which is based on aggregated information from a virtual routing plane using the BGP protocol and a folded-Clos physical network topology based on OpenFlow 1.3 devices. The experimental evaluation shows that the proposed algorithm performs efficient load balancing on the data center network and altogether yields better utilization of the physical resources. The proposed bandwidth allocation strategy exhibits simplicity and flexibility to attend different traffic communication patterns while yielding an elastic load balanced network. Finally, we argue that the algorithm and the architecture proposed can be extended to achieve performance, scalability and many other features required in data center network architectures / Mestrado / Ciência da Computação / Mestre em Ciência da Computação