No description available.

No description available.

In this thesis, we propose a new protocol named Layered TCP with RTT Compensation (LTCP-RC, for short). LTCP-RC is a simple modification to the congestion window response of the high-speed protocol, Layered TCP (LTCP). In networks characterized by large link delays and high RTTs, LTCP-RC makes the LTCP protocol more scalable. Ack-clocked schemes, similar to TCP, suffer performance problems like long convergence time and throughput degradation, when RTT experienced by the flow increases. Also, when flows with different RTTs compete, the problem of unfairness among competing flows becomes worse in the case of high-speed protocols. LTCP-RC uses an RTT Compensation technique in order to solve these problems. This thesis presents a general framework to decide the function for RTT Compensation factor and two particular design choices are analyzed in detail. The first algorithm uses a fixed function based on the minimum RTT observed by the flow. The second algorithm uses an adaptive scheme which regulates itself according to the dynamic network conditions. Evaluation of the performance of these schemes is done using analysis and ns-2 simulations. LTCP-RC exhibits significant performance improvement in terms of reduced convergence time, low drop rates, increased utilization in presence of links with channel errors and good fairness properties between the flows,. The scheme is simple to understand, easy to implement on the TCP/IP stack and does not require any additional support from the network resources. The choice of parameters can be influenced to tune the RTT unfairness of the scheme, which is not possible in TCP or other high-speed protocols. The flexible nature of the analysis framework has laid the ground work for the development of new schemes, which can improve the performance of the window based protocols in high delay and heterogeneous networks.

Reliable and efficient data transfer on the Internet is an important issue. Since late 70's the protocol responsible for that has been the de facto standard TCP, which has proven to be successful through out the years, its self-managed congestion control algorithms have retained the stability of the Internet for decades. However, the variety of existing new technologies such as high-speed networks (e.g. fibre optics) with high-speed long-delay set-up (e.g. cross-Atlantic links) and wireless technologies have posed lots of challenges to TCP congestion control algorithms. The congestion control research community proposed solutions to most of these challenges. This dissertation adds to the existing work by: firstly tackling the highspeed long-delay problem of TCP, we propose enhancements to one of the existing TCP variants (part of Linux kernel stack). We then propose our own variant: TCP-Gentle. Secondly, tackling the challenge of differentiating the wireless loss from congestive loss in a passive way and we propose a novel loss differentiation algorithm which quantifies the noise in packet inter arrival times and use this information together with the span (ratio of maximum to minimum packet inter arrival times) to adapt the multiplicative decrease factor according to a predefined logical formula. Finally, extending the well-known drift model of TCP to account for wireless loss and some hypothetical cases (e.g. variable multiplicative decrease), we have undertaken stability analysis for the new version of the model.

La tendance en HPC est à l'accroissement du nombre de coeurs par noeud de calcul pour une quantité totale de mémoire par noeud constante. A large échelle, l'un des principaux défis pour les applications parallèles est de garder une faible consommation mémoire. Cette thèse présente une couche de communication multi-threadée sur Infiniband, laquelle fournie de bonnes performances et une faible consommation mémoire. Nous ciblons les applications scientifiques parallélisées grâce à la bibliothèque MPI ou bien combinées avec un modèle de programmation en mémoire partagée. En partant du constat que le nombre de connexions réseau et de buffers de communication est critique pour la mise à l'échelle des bibliothèques MPI, la première contribution propose trois approches afin de contrôler leur utilisation. Nous présentons une topologie virtuelle extensible et entièrement connectée pour réseaux rapides orientés connexion. Dans un contexte agrégeant plusieurs cartes permettant d'ajuster dynamiquement la configuration des buffers réseau utilisant la technologie RDMA. La seconde contribution propose une optimisation qui renforce le potentiel d'asynchronisme des applications MPI, laquelle montre une accélération de deux des communications. La troisième contribution évalue les performances de plusieurs bibliothèques MPI exécutant une application de modélisation sismique en contexte hybride. Les expériences sur des noeuds de calcul jusqu'à 128 coeurs montrent une économie de 17 % sur la mémoire. De plus, notre couche de communication multi-threadée réduit le temps d'exécution dans le cas où plusieurs threads OpenMP participent simultanément aux communications MPI. / A recent trend in high performance computing shows a rising number of cores per compute node, while the total amount of memory per compute node remains constant. To scale parallel applications on such large machines, one of the major challenges is to keep a low memory consumption. This thesis develops a multi-threaded communication layer over Infiniband which provides both good performance of communications and a low memory consumption. We target scientific applications parallelized using the MPI standard in pure mode or combined with a shared memory programming model. Starting with the observation that network endpoints and communication buffers are critical for the scalability of MPI runtimes, the first contribution proposes three approaches to control their usage. We introduce a scalable and fully-connected virtual topology for connection-oriented high-speed networks. In the context of multirail configurations, we then detail a runtime technique which reduces the number of network connections. We finally present a protocol for dynamically resizing network buffers over the RDMA technology. The second contribution proposes a runtime optimization to enforce the overlap potential of MPI communications, showing a 2x improvement factor on communications. The third contribution evaluates the performance of several MPI runtimes running a seismic modeling application in a hybrid context. On large compute nodes up to 128 cores, the introduction of OpenMP in the MPI application saves up to 17 % of memory. Moreover, we show a performance improvement with our multi-threaded communication layer where the OpenMP threads concurrently participate to the MPI communications

Orientador: Nelson Luis Saldanha da Fonseca / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-13T20:11:45Z (GMT). No. of bitstreams: 1 Michel_NeilaFernanda_M.pdf: 1961141 bytes, checksum: f4470a332c9b5f764b7bd3b912c27db1 (MD5) Previous issue date: 2008 / Resumo: O protoloco TCP Reno começou a operar em 01 de Janeiro de 1983. Desde sua concepção, houveram expressivos avanços nas tecnologias de transmissão de dados, como a introdução de enlaces ópticos e de satélites, 'que permitem altas taxas de transmissão. O TCP Reno continua sendo o protocolo padrão na Internet. Entretanto existem inúmeras dificuldades na sua operação, que causam baixo desempenho em redes com produto banda-atraso elevado, devido ao mecanismo de atualização do tamanho da janela de congestionamento. Para superar suas deficiências, vários protocolos variantes do TCP Reno têm sido propostos com o objetivo de melhorar a utilização de recursos em redes de alta velocidade, e por esta razão são denominados "Protocolos para Redes de Alta Velocidade" . Para que as proposições de variantes do TCP levem a protocolos que operem eficientemente em redes com alto valor do produto banda-atraso, é necessário que se avalie quais variações da dinâmica do crescimento da janela de transmissão são efetivas. Nesta dissertação avalia-se um conjunto de protocolos variantes do TCP para redes de alta velocidade através de simulações. Diversas propriedades importantes para protocolos do tipo TCP são investigadas / Abstract: The importance of the Reno TCP protocol started operating in January 01 st, 1983. Since this milestone, there were expressive advances on data transmission technologies, such as those of optical fibers and satellite channels which allow high transmission rates on long distances. Reno TCP remains the Internet standard protocol. However, its operation presents several dificulties in networks with high bandwidth-delay product, producing poor performance due its window congestion mechanism. In order to overcome its deficiencies, several variants have been proposed with the aim of enhancing its utilization in high speed networks. These variants are generally called "High Speed Networks Protocols". Transmission chàracteristics, properties and performance evaluation are essencial to the identification of problems and avoidance of poor performance. In this dissertation a set of TCP variants protocols for high speed networks is evaluated. Several important properties of these protocols are investigated / Mestrado / Redes de Computadores / Mestre em Ciência da Computação

Orientadores: Michel Zamboni Rached, Julio Cesar Rodrigues Fernandes de Oliveira / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-23T22:08:51Z (GMT). No. of bitstreams: 1 Silva_EdsonPortoda_M.pdf: 2920023 bytes, checksum: 44d415ca69c358709f1b816f7e603047 (MD5) Previous issue date: 2013 / Resumo: O Resumo poderá ser visualizado no texto completo da tese digital / Abstract: The complete Abstract is available with the full electronic document. / Mestrado / Telecomunicações e Telemática / Mestre em Engenharia Elétrica

Orientador: Nelson Luis Saldanha da Fonseca / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-19T17:25:04Z (GMT). No. of bitstreams: 1 Froldi_CarlosAugusto_M.pdf: 543951 bytes, checksum: e4f64a2921c88fd1461445bd39a04278 (MD5) Previous issue date: 2011 / Resumo: Os protocolos da camada de transporte na Internet, TCP e UDP, não oferecem serviços para transmissão eficiente de fluxos multimídia, porém, este último é adotado com maior freqüência por essas aplicações. Uma proposta de um novo protocolo da camada de transporte, chamado DCCP, foi elaborada para atender a demanda das aplicações multimídia e substituir o protocolo UDP. A presente dissertação propõe uma variante deste protocolo, chamada Fast DCCP, para operar de maneira eficiente em redes de alta velocidade. A nova variante foi avaliada através de experimentos de simulação, utilizando o NS-2 e experimentos de medição, utilizando o sistema operacional Linux / Abstract: The Internet transport layer protocols, TCP and UDP, do not provide efficient transport service for multimedia streams. UDP is usually used for these applications, due to its low overhead. A new transport layer protocol, called DCCP, was proposed to meet the demand of multimedia applications, aiming at replacing the UDP protocol. This dissertation will propose a variant for this protocol, called Fast DCCP, for operating efficiently on high-speed networks. It was evaluated by simulation using the NS-2 network simulator and measurements of the FAST DCCP protocol operation in the Linux operating system / Mestrado / Ciência da Computação / Mestre em Ciência da Computação

Orientadores: Aldário Chrestani Bordonalli, Júlio César Rodrigues Fernandes de Oliveira / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-22T02:02:14Z (GMT). No. of bitstreams: 1 Ribeiro_VitorBedotti_M.pdf: 8224184 bytes, checksum: f63b4dc7f1e4170cdcb48f49bf1b4155 (MD5) Previous issue date: 2012 / Resumo: As mudanças nos padrões de tráfego na internet forçam a constante evolução das redes ópticas, que, por serem sistemas com grande capacidade e se estenderam por milhares de quilômetros, são a principal aposta para suprir a demanda por banda. A solução adotada foi aumentar a eficiência espectral da transmissão, por meio do uso de formatos de modulação de alta ordem, como o PSK e o QAM, da detecção coerente e da multiplexação em polarização. Nessa dissertação, os conceitos de uma transmissão coerente digital e os principais algoritmos para a recepção e recuperação de sinal são apresentados, considerando o formato de modulação DP-QPSK com taxa agregada de 112 Gb/s. É proposto, também, um método de adaptação de ganho para o algoritmo de módulo constante (CMA), que acelera a convergência e torna a rotina que rastreia as mudanças no estado de polarização do sinal óptico mais rápido e robusto. Os algoritmos para cada funcionalidade foram testados individualmente em simulação. O melhor conjunto de algoritmos é analisado e configurações experimentais em back-to-back com carregamento de ruído ASE, em anel de recirculação óptico de 225 km com 80 canais DWDM e, também, em transmissão pela Rede Experimental de Alta Velocidade GIGA no trecho Campinas - São Paulo - Campinas. Para a configuração em back-to-back, o limite do sistema foi de 10,5 dB de OSNR. No anel de recirculação, foi possível a transmissão de 80 canais a 112 Gb/s, totalizando 8,96 Tbit/s por 2925 km e ocupando 50 GHz de banda, ou por 2475 km e ocupando 25 GHz. Na rede GIGA, o sistema se mostrou estável, com taxa de erro de bits bem inferior ao limite por mais de 3 horas de análise. O método de adaptação de ganho proposto conseguiu acompanhar mais de 20 Mrad/s e 3 Mrad/s de variação angular no estado de polarização para 34 dB e 15 dB de OSNR, respectivamente, enquanto o CMA sem adaptação de ganho falha a 10 Mrad/s e 1,5 Mrad/s nas mesmas condições. Os resultados demonstram a eficácia e a estabilidade dos algoritmos e a capacidade de recuperação do sinal em simulação, em experimentos laboratoriais ou em campo / Abstract: The changes in traffic patterns due to the Internet force the constant evolution of optical networks. These systems, designed to deliver high transmission capacity over thousands of kilometers, are the key solution to meet the demand for bandwidth. The recent tendency to provide wider bandwidth without changing the infrastructure already deployed has been to increase the spectral efficiency of transmission by using high order modulation formats, such as PSK and QAM, and employment coherent detection along with polarization multiplexing. In this work the concepts of optical digital coherent systems and the main algorithms for signal reception and recovery are presented, considering QPSK modulation format and polarization multiplexing (DP-QPSK) at 112 Gb/s. It is also proposed a method for adaptive adjustment of the constant modulus algorithm (CMA) gain, which accelerates the convergence and makes the algorithm faster and more robust to changes in the state of polarization of the optical signal. The algorithms for each function were tested individually in a simulation setup. The best set of algorithms were then obtained and applied in the receiver end of different experimental setups: back-to-back with ASE noise loading; 225-km optical recirculation loop with 80 DWDM channels; and a field transmission over part of the high-capacity experimental network GIGA (Campinas - São Paulo - Campinas). For the back-to-back setup, the system limit was achieved at 10.5 dB of OSNR. For the recirculation loop, it was possible to transmit up to 80 channels at 112 Gb/s, or 8.96 Tbit/s for overall system capacity, over 2925 km, considering 50-GHz bandwidth signal, or 2475 km for 25-GHz bandwidth signal. For the GIGA network, the system was stable with bit error rate well below the FEC threshold during the 3-hour analysis. The gain adaptation method proposed was able to track over 20 Mrad/s and 3 Mrad/s of angular rotation rate in the state of polarization for OSNRs of 34 dB and 15 dB, respectively, while the CMA without gain adaptation fails to track frequencies above 10 Mrad/s and 1.5 Mrad/s in the same conditions. The results show that the set of algorithms is effective, stable and capable of recovering the signal during simulation, laboratory experiments or field trials / Mestrado / Telecomunicações e Telemática / Mestre em Engenharia Elétrica

The master thesis discusses about the relation between high-speed networks and intelligent system installation features and possibilities of this mutual interaction. Furthermore principles of KNX bus tunneling through IP networks are theoretically analyzed (KNXnet/IP). Practical use of implements KNXnet/IP shows real elements on the test panel. The control is realized by KNX@Home in the Ubuntu environment. Simulation of the real state tunneling KNX bus represent model in Opnet Modeler, which shows how the delays may change due to network load.