Spelling suggestions: "subject:"datacenter networks"" "subject:"datacenters networks""
1 |
Achieving predictable, guaranted and work-conserving performance in datacenter networks / Atingindo desempenho previsivel, garantido e com conservação de trabalhos em redes datacenterMarcon, Daniel Stefani January 2017 (has links)
A interferência de desempenho é um desafio bem conhecido em redes de datacenter (DCNs), permanecendo um tema constante de discussão na literatura. Diversos estudos concluíram que a largura de banda disponível para o envio e recebimento de dados entre máquinas virtuais (VMs) pode variar por um fator superior a cinco, resultando em desempenho baixo e imprevisível para as aplicações. Trabalhos na literatura têm proposto técnicas que resultam em subutilização de recursos, introduzem sobrecarga de gerenciamento ou consideram somente recursos de rede. Nesta tese, são apresentadas três propostas para lidar com a interferência de desempenho em DCNs: IoNCloud, Predictor e Packer. O IoNCloud está baseado na observação que diferentes aplicações não possuem pico de damanda de banda ao mesmo tempo. Portanto, ele busca prover desempenho previsível e garantido enquanto minimiza a subutilização dos recursos de rede. Isso é alcançado por meio (a) do agrupamento de aplicações (de acordo com os seus requisitos temporais de banda) em redes virtuais (VNs); e (b) da alocação dessas VNs no substrato físico. Apesar de alcançar os seus objetivos, ele não provê conservação de trabalho entre VNs, o que limita a utilização de recursos ociosos. Nesse contexto, o Predictor, uma evolução do IoNCloud, programa dinamicamente a rede em DCNs baseadas em redes definidas por software (SDN) e utiliza dois novos algoritmos para prover garantias de desempenho de rede com conservação de trabalho. Além disso, ele foi projetado para ser escalável, considerando o número de regras em tabelas de fluxo e o tempo de instalação das regras para um novo fluxo em DCNs com milhões de fluxos ativos. Apesar dos benefícios, o IoNCloud e o Predictor consideram apenas os recursos de rede no processo de alocação de aplicações na infraestrutura física. Isso leva à fragmentação de outros tipos de recursos e, consequentemente, resulta em um menor número de aplicações sendo alocadas. O Packer, em contraste, busca prover desempenho de rede previsível e garantido e minimizar a fragmentação de diferentes tipos de recursos. Estendendo a observação feita ao IoNCloud, a observação-chave é que as aplicações têm demandas complementares ao longo do tempo para múltiplos recursos. Desse modo, o Packer utiliza (i) uma nova abstração para especificar os requisitos temporais das aplicações, denominada TI-MRA (Time- Interleaved Multi-Resource Abstraction); e (ii) uma nova estratégia de alocação de recursos. As avaliações realizadas mostram os benefícios e as sobrecargas do IoNCloud, do Predictor e do Packer. Em particular, os três esquemas proveem desempenho de rede previsível e garantido; o Predictor reduz o número de regras OpenFlow em switches e o tempo de instalação dessas regras para novos fluxos; e o Packer minimiza a fragmentação de múltiplos tipos de recursos. / Performance interference has been a well-known problem in datacenter networks (DCNs) and one that remains a constant topic of discussion in the literature. Several measurement studies concluded that throughput achieved by virtual machines (VMs) in current datacenters can vary by a factor of five or more, leading to poor and unpredictable overall application performance. Recent efforts have proposed techniques that present some shortcomings, such as underutilization of resources, significant management overhead or negligence of non-network resources. In this thesis, we introduce three proposals that address performance interference in DCNs: IoNCloud, Predictor and Packer. IoNCloud leverages the key observation that temporal bandwidth demands of cloud applications do not peak at exactly the same time. Therefore, it seeks to provide predictable and guaranteed performance while minimizing network underutilization by (a) grouping applications in virtual networks (VNs) according to their temporal network usage and need of isolation; and (b) allocating these VNs on the cloud substrate. Despite achieving its objective, IoNCloud does not provide work-conserving sharing among VNs, which limits utilization of idle resources. Predictor, an evolution over IoNCloud, dynamically programs the network in Software-Defined Networking (SDN)-based DCNs and uses two novel algorithms to provide network guarantees with work-conserving sharing. Furthermore, Predictor is designed with scalability in mind, taking into consideration the number of entries required in flow tables and flow setup time in DCNs with high turnover and millions of active flows. IoNCloud and Predictor neglect resources other than the network at allocation time. This leads to fragmentation of non-network resources and, consequently, results in less applications being allocated in the infrastructure. Packer, in contrast, aims at providing predictable and guaranteed network performance while minimizing overall multi-resource fragmentation. Extending the observation presented for IoNCloud, the key insight for Packer is that applications have complementary demands across time for multiple resources. To enable multi-resource allocation, we devise (i) a new abstraction for specifying temporal application requirements (called Time-Interleaved Multi-Resource Abstraction – TI-MRA); and (ii) a new allocation strategy. We evaluated IoNCloud, Predictor and Packer, showing their benefits and overheads. In particular, all of them provide predictable and guaranteed network performance; Predictor reduces flow table size in switches and flow setup time; and Packer minimizes multi-resource fragmentation.
|
2 |
Achieving predictable, guaranted and work-conserving performance in datacenter networks / Atingindo desempenho previsivel, garantido e com conservação de trabalhos em redes datacenterMarcon, Daniel Stefani January 2017 (has links)
A interferência de desempenho é um desafio bem conhecido em redes de datacenter (DCNs), permanecendo um tema constante de discussão na literatura. Diversos estudos concluíram que a largura de banda disponível para o envio e recebimento de dados entre máquinas virtuais (VMs) pode variar por um fator superior a cinco, resultando em desempenho baixo e imprevisível para as aplicações. Trabalhos na literatura têm proposto técnicas que resultam em subutilização de recursos, introduzem sobrecarga de gerenciamento ou consideram somente recursos de rede. Nesta tese, são apresentadas três propostas para lidar com a interferência de desempenho em DCNs: IoNCloud, Predictor e Packer. O IoNCloud está baseado na observação que diferentes aplicações não possuem pico de damanda de banda ao mesmo tempo. Portanto, ele busca prover desempenho previsível e garantido enquanto minimiza a subutilização dos recursos de rede. Isso é alcançado por meio (a) do agrupamento de aplicações (de acordo com os seus requisitos temporais de banda) em redes virtuais (VNs); e (b) da alocação dessas VNs no substrato físico. Apesar de alcançar os seus objetivos, ele não provê conservação de trabalho entre VNs, o que limita a utilização de recursos ociosos. Nesse contexto, o Predictor, uma evolução do IoNCloud, programa dinamicamente a rede em DCNs baseadas em redes definidas por software (SDN) e utiliza dois novos algoritmos para prover garantias de desempenho de rede com conservação de trabalho. Além disso, ele foi projetado para ser escalável, considerando o número de regras em tabelas de fluxo e o tempo de instalação das regras para um novo fluxo em DCNs com milhões de fluxos ativos. Apesar dos benefícios, o IoNCloud e o Predictor consideram apenas os recursos de rede no processo de alocação de aplicações na infraestrutura física. Isso leva à fragmentação de outros tipos de recursos e, consequentemente, resulta em um menor número de aplicações sendo alocadas. O Packer, em contraste, busca prover desempenho de rede previsível e garantido e minimizar a fragmentação de diferentes tipos de recursos. Estendendo a observação feita ao IoNCloud, a observação-chave é que as aplicações têm demandas complementares ao longo do tempo para múltiplos recursos. Desse modo, o Packer utiliza (i) uma nova abstração para especificar os requisitos temporais das aplicações, denominada TI-MRA (Time- Interleaved Multi-Resource Abstraction); e (ii) uma nova estratégia de alocação de recursos. As avaliações realizadas mostram os benefícios e as sobrecargas do IoNCloud, do Predictor e do Packer. Em particular, os três esquemas proveem desempenho de rede previsível e garantido; o Predictor reduz o número de regras OpenFlow em switches e o tempo de instalação dessas regras para novos fluxos; e o Packer minimiza a fragmentação de múltiplos tipos de recursos. / Performance interference has been a well-known problem in datacenter networks (DCNs) and one that remains a constant topic of discussion in the literature. Several measurement studies concluded that throughput achieved by virtual machines (VMs) in current datacenters can vary by a factor of five or more, leading to poor and unpredictable overall application performance. Recent efforts have proposed techniques that present some shortcomings, such as underutilization of resources, significant management overhead or negligence of non-network resources. In this thesis, we introduce three proposals that address performance interference in DCNs: IoNCloud, Predictor and Packer. IoNCloud leverages the key observation that temporal bandwidth demands of cloud applications do not peak at exactly the same time. Therefore, it seeks to provide predictable and guaranteed performance while minimizing network underutilization by (a) grouping applications in virtual networks (VNs) according to their temporal network usage and need of isolation; and (b) allocating these VNs on the cloud substrate. Despite achieving its objective, IoNCloud does not provide work-conserving sharing among VNs, which limits utilization of idle resources. Predictor, an evolution over IoNCloud, dynamically programs the network in Software-Defined Networking (SDN)-based DCNs and uses two novel algorithms to provide network guarantees with work-conserving sharing. Furthermore, Predictor is designed with scalability in mind, taking into consideration the number of entries required in flow tables and flow setup time in DCNs with high turnover and millions of active flows. IoNCloud and Predictor neglect resources other than the network at allocation time. This leads to fragmentation of non-network resources and, consequently, results in less applications being allocated in the infrastructure. Packer, in contrast, aims at providing predictable and guaranteed network performance while minimizing overall multi-resource fragmentation. Extending the observation presented for IoNCloud, the key insight for Packer is that applications have complementary demands across time for multiple resources. To enable multi-resource allocation, we devise (i) a new abstraction for specifying temporal application requirements (called Time-Interleaved Multi-Resource Abstraction – TI-MRA); and (ii) a new allocation strategy. We evaluated IoNCloud, Predictor and Packer, showing their benefits and overheads. In particular, all of them provide predictable and guaranteed network performance; Predictor reduces flow table size in switches and flow setup time; and Packer minimizes multi-resource fragmentation.
|
3 |
Achieving predictable, guaranted and work-conserving performance in datacenter networks / Atingindo desempenho previsivel, garantido e com conservação de trabalhos em redes datacenterMarcon, Daniel Stefani January 2017 (has links)
A interferência de desempenho é um desafio bem conhecido em redes de datacenter (DCNs), permanecendo um tema constante de discussão na literatura. Diversos estudos concluíram que a largura de banda disponível para o envio e recebimento de dados entre máquinas virtuais (VMs) pode variar por um fator superior a cinco, resultando em desempenho baixo e imprevisível para as aplicações. Trabalhos na literatura têm proposto técnicas que resultam em subutilização de recursos, introduzem sobrecarga de gerenciamento ou consideram somente recursos de rede. Nesta tese, são apresentadas três propostas para lidar com a interferência de desempenho em DCNs: IoNCloud, Predictor e Packer. O IoNCloud está baseado na observação que diferentes aplicações não possuem pico de damanda de banda ao mesmo tempo. Portanto, ele busca prover desempenho previsível e garantido enquanto minimiza a subutilização dos recursos de rede. Isso é alcançado por meio (a) do agrupamento de aplicações (de acordo com os seus requisitos temporais de banda) em redes virtuais (VNs); e (b) da alocação dessas VNs no substrato físico. Apesar de alcançar os seus objetivos, ele não provê conservação de trabalho entre VNs, o que limita a utilização de recursos ociosos. Nesse contexto, o Predictor, uma evolução do IoNCloud, programa dinamicamente a rede em DCNs baseadas em redes definidas por software (SDN) e utiliza dois novos algoritmos para prover garantias de desempenho de rede com conservação de trabalho. Além disso, ele foi projetado para ser escalável, considerando o número de regras em tabelas de fluxo e o tempo de instalação das regras para um novo fluxo em DCNs com milhões de fluxos ativos. Apesar dos benefícios, o IoNCloud e o Predictor consideram apenas os recursos de rede no processo de alocação de aplicações na infraestrutura física. Isso leva à fragmentação de outros tipos de recursos e, consequentemente, resulta em um menor número de aplicações sendo alocadas. O Packer, em contraste, busca prover desempenho de rede previsível e garantido e minimizar a fragmentação de diferentes tipos de recursos. Estendendo a observação feita ao IoNCloud, a observação-chave é que as aplicações têm demandas complementares ao longo do tempo para múltiplos recursos. Desse modo, o Packer utiliza (i) uma nova abstração para especificar os requisitos temporais das aplicações, denominada TI-MRA (Time- Interleaved Multi-Resource Abstraction); e (ii) uma nova estratégia de alocação de recursos. As avaliações realizadas mostram os benefícios e as sobrecargas do IoNCloud, do Predictor e do Packer. Em particular, os três esquemas proveem desempenho de rede previsível e garantido; o Predictor reduz o número de regras OpenFlow em switches e o tempo de instalação dessas regras para novos fluxos; e o Packer minimiza a fragmentação de múltiplos tipos de recursos. / Performance interference has been a well-known problem in datacenter networks (DCNs) and one that remains a constant topic of discussion in the literature. Several measurement studies concluded that throughput achieved by virtual machines (VMs) in current datacenters can vary by a factor of five or more, leading to poor and unpredictable overall application performance. Recent efforts have proposed techniques that present some shortcomings, such as underutilization of resources, significant management overhead or negligence of non-network resources. In this thesis, we introduce three proposals that address performance interference in DCNs: IoNCloud, Predictor and Packer. IoNCloud leverages the key observation that temporal bandwidth demands of cloud applications do not peak at exactly the same time. Therefore, it seeks to provide predictable and guaranteed performance while minimizing network underutilization by (a) grouping applications in virtual networks (VNs) according to their temporal network usage and need of isolation; and (b) allocating these VNs on the cloud substrate. Despite achieving its objective, IoNCloud does not provide work-conserving sharing among VNs, which limits utilization of idle resources. Predictor, an evolution over IoNCloud, dynamically programs the network in Software-Defined Networking (SDN)-based DCNs and uses two novel algorithms to provide network guarantees with work-conserving sharing. Furthermore, Predictor is designed with scalability in mind, taking into consideration the number of entries required in flow tables and flow setup time in DCNs with high turnover and millions of active flows. IoNCloud and Predictor neglect resources other than the network at allocation time. This leads to fragmentation of non-network resources and, consequently, results in less applications being allocated in the infrastructure. Packer, in contrast, aims at providing predictable and guaranteed network performance while minimizing overall multi-resource fragmentation. Extending the observation presented for IoNCloud, the key insight for Packer is that applications have complementary demands across time for multiple resources. To enable multi-resource allocation, we devise (i) a new abstraction for specifying temporal application requirements (called Time-Interleaved Multi-Resource Abstraction – TI-MRA); and (ii) a new allocation strategy. We evaluated IoNCloud, Predictor and Packer, showing their benefits and overheads. In particular, all of them provide predictable and guaranteed network performance; Predictor reduces flow table size in switches and flow setup time; and Packer minimizes multi-resource fragmentation.
|
4 |
Software-defined Buffer Management and Robust Congestion Control for Modern Datacenter NetworksDanushka N Menikkumbura (12208121) 20 April 2022 (has links)
<p> Modern datacenter network applications continue to demand ultra low latencies and very high throughputs. At the same time, network infrastructure keeps achieving higher speeds and larger bandwidths. We still need better network management solutions to keep these two demand and supply fronts go hand-in-hand. There are key metrics that define network performance such as flow completion time (the lower the better), throughput (the higher the better), and end-to-end latency (the lower the better) that are mainly governed by how effectively network application get their fair share of network resources. We observe that buffer utilization on network switches gives a very accurate indication of network performance. Therefore, network buffer management is important in modern datacenter networks, and other network management solutions can be efficiently built around buffer utilization. This dissertation presents three solutions based on buffer use on network switches.</p>
<p> This dissertation consists of three main sections. The first section is on a specification language for buffer management in modern programmable switches. The second section is on a congestion control solution for Remote Direct Memory Access (RDMA) networks. The third section is on a solution to head-of-the-line blocking in modern datacenter networks.</p>
|
5 |
Data center optical networks : short- and long-term solutions / Réseaux optiques pour les centres de données : solutions à court et long termeMestre Adrover, Miquel Angel 21 October 2016 (has links)
Les centres de données deviennent de plus en plus importants, allant de petites fermes de serveurs distribuées à des grandes fermes dédiées à des tâches spécifiques. La diffusion de services "dans le nuage" conduit à une augmentation incessante de la demande de trafic dans les centres de données. Dans cette thèse, nous étudions l'évolution des réseaux dans les centres de données et proposons des solutions à court et à long terme pour leur intra-connexion physique. Aujourd'hui, la croissance de la demande de trafic met en lumière la nécessité urgente d’interfaces à grande vitesse capables de faire face à la bande passante exigeant de nouvelles applications. Ainsi, à court terme, nous proposons de nouveaux transpondeurs optiques à haut débit, mais à faible coût, permettant la transmission de 200 Gb /s utilisant des schémas de modulation en intensité et à détection directe. Plusieurs types de modulations d’impulsions en amplitude avancées sont explorés, tout en augmentant la vitesse à des débits symboles allant jusqu’à 100 GBd. La génération électrique à haute vitesse est réalisé grâce à un nouveau convertisseur analogique-numérique intégré, capable de doubler les vitesses des entrées et de générer des signaux à plusieurs niveaux d’amplitude. Cependant, le trafic continuera sa croissance. Les centres de données actuels reposent sur plusieurs niveaux de commutateurs électroniques pour construire un réseau d'interconnexion capable de supporter une telle grande quantité de trafic. Dans une telle architecture, la croissance du trafic est directement liée à une augmentation du nombre des composants du réseau, y-compris les commutateurs avec plus de ports, les interfaces et les câbles. Le coût et la consommation d'énergie qui peut être attendus à l'avenir est intenable, ce qui appelle à une réévaluation du réseau. Par conséquent, nous présentons ensuite un nouveau concept fondé sur la commutation de "slots" optiques (Burst Optical Slot Switching, i.e. BOSS) dans lequel les serveurs sont connectés via des nœuds BOSS à travers des anneaux de fibres multiplexé en longueur d'onde et en temps, et organisés dans une topologie en tore. Au cours de cette thèse, nous étudions la mise en œuvre des nœuds BOSS; en particulier, la matrice de commutation et les transpondeurs optiques. L'élément principal au sein de la matrice de commutation est le bloqueur de slots, qui est capable d'effacer n’importe quel paquet (slot) sur n’importe quelle longueur d'onde en quelques nanosecondes seulement. D'une part, nous explorons l'utilisation d'amplificateurs optiques à semi-conducteurs comme portes optiques à utiliser dans le bloqueur des slots, et étudier leur cascade. D'autre part, nous développons un bloqueur de slots intégré monolithiquement capable de gérer jusqu'à seize longueurs d'onde avec la diversité de polarisation. Ensuite, nous présentons plusieurs architectures de transpondeur et nous étudions leur performance. La signalisation des transpondeurs doit répondre à deux exigences principales: le fonctionnement en mode paquet et la résistance au filtrage serré. D'abord, nous utilisons des transpondeurs élastiques qui utilisent des modulations Nyquist N-QAM, et qui adaptent le format de modulation en fonction du nombre de nœuds à traverser. Ensuite, nous proposons l'utilisation du multiplexage par répartition orthogonale de la fréquence en cohérence optique (CO-OFDM). Avec une structure de paquet inhérente et leur grande adaptabilité fréquentielle, nous démontrons que les transpondeurs CO-OFDM offrent une capacité plus élevée et une meilleure portée que leurs homologues Nyquist. Finalement, nous comparons notre solution BOSS avec la topologie Clos replié utilisée aujourd'hui. Nous montrons que notre architecture BOSS nécessite 400 fois moins de transpondeurs et de câbles que les réseaux de commutation électronique d'aujourd'hui, ce qui ouvre la voie à des centres de données hautement évolutifs et durables / Data centers are becoming increasingly important and ubiquitous, ranging from large server farms dedicated to various tasks such as data processing, computing, data storage or the combination thereof, to small distributed server farms. The spread of cloud services is driving a relentless increase of traffic demand in datacenters, which is doubling every 12 to 15 months. Along this thesis we study the evolution of data center networks and present short- and long-term solutions for their physical intra-connection. Today, rapidly-growing traffic in data centers spotlights the urgent need for high-speed low-cost interfaces capable to cope with hungry-bandwidth demanding new applications. Thereby, in the short-term we propose novel high-datarate low-cost optical transceivers enabling up to 200 Gb/s transmission using intensity-modulation and direct-detection schemes. Several advanced pulse amplitude modulation schemes are explored while increasing speeds towards record symbol-rates, as high as 100 GBd. High-speed electrical signaling is enabled by an integrated selector-power digital-to- analog converter, capable of doubling input baud-rates while outputting advance multi-level pulse amplitude modulations. Notwithstanding, data centers’ global traffic will continue increasing incessantly. Current datacenters rely on high-radix all-electronic Ethernet switches to build an interconnecting network capable to pave with such vast amount of traffic. In such architecture, traffic growth directly relates to an increase of networking components, including switches with higher port-count, interfaces and cables. Unsustainable cost and energy consumption that can be expected in the future calls for a network reassessment. Therefore, we subsequently present a novel concept for intra-datacenter networks called burst optical slot switching (BOSS); in which servers are connected via BOSS nodes through wavelength- and time-division multiplexed fiber rings organized in a Torus topology. Along this thesis we investigate on the implementation of BOSS nodes; in particular, the switching fabric and the optical transceivers. The main element within the switching fabric is the slot blocker, which is capable of erasing any packet of any wavelength in a nanosecond time-scale. On the one hand, we explore the use of semiconductor optical amplifiers as means of gating element to be used within the slot blocker and study their cascadability. On the other hand we develop a monolithically integrated slot blocker capable of handling up to sixteen wavelength channels with dual-polarization diversity. Then we present several transceiver architectures and study their performances. Transceivers’ signaling needs to fulfill two main requirements: packet-mode operation, i.e. being capable of recovering few microsecond –long bursts; and resiliency to tight filtering, which occurs when cascading many nodes (e.g. up to 100). First we build packet-mode Nyquist-pulse-shaped N-QAM transceivers, which adapt the modulation format as a function of the number of nodes to traverse. Later we propose the use of coherent-optical orthogonal frequency division multiplexing (CO-OFDM). With inherent packet structure and high spectral tailoring capabilities, we demonstrate that CO-OFDM-based transceivers offer higher capacity and enhanced reach than its Nyquist counterpart. Finally, we compare our BOSS solution to today’s Folded Clos topology, and show that our BOSS architecture requires x400 fewer transponders and cables than today’s electronic switching networks, which paves the way to highly scalable and sustainable datacenters
|
Page generated in 0.0891 seconds