The growing computational demand of scientific applications has greatly motivated the development of large-scale high-performance computing (HPC) systems in the past decade. To accommodate the increasing demand of applications, HPC systems have been going through dramatic architectural changes (e.g., introduction of many-core and multi-core systems, rapid growth of complex interconnection network for efficient communication between thousands of nodes), as well as significant increase in size (e.g., modern supercomputers consist of hundreds of thousands of nodes). With such changes in architecture and size, the energy consumption by these systems has increased significantly. With the advent of exascale supercomputers in the next few years, power consumption of the HPC systems will surely increase; some systems may even consume hundreds of megawatts of electricity. Demand response programs are designed to help the energy service providers to stabilize the power system by reducing the energy consumption of participating systems during the time periods of high demand power usage or temporary shortage in power supply. This dissertation focuses on developing energy-efficient demand-response models and algorithms to enable HPC system's demand response participation. In the first part, we present interconnection network models for performance prediction of large-scale HPC applications. They are based on interconnected topologies widely used in HPC systems: dragonfly, torus, and fat-tree. Our interconnect models are fully integrated with an implementation of message-passing interface (MPI) that can mimic most of its functions with packet-level accuracy. Extensive experiments show that our integrated models provide good accuracy for predicting the network behavior, while at the same time allowing for good parallel scaling performance. In the second part, we present an energy-efficient demand-response model to reduce HPC systems' energy consumption during demand response periods. We propose HPC job scheduling and resource provisioning schemes to enable HPC system's emergency demand response participation. In the final part, we propose an economic demand-response model to allow both HPC operator and HPC users to jointly reduce HPC system's energy cost. Our proposed model allows the participation of HPC systems in economic demand-response programs through a contract-based rewarding scheme that can incentivize HPC users to participate in demand response.

Field-Programmable Gate Arrays (FPGAs) systems now comprise many processing elements that are processors running software and hardware engines used to accelerate specific functions. To make the programming of such a system simpler, it is easiest to think of a shared-memory environment, much like in current multi-core processor systems. This thesis introduces a novel, shared-memory, cache-coherent infrastructure for heterogeneous systems implemented on FPGAs that can then form the basis of a shared-memory programming model for heterogeneous systems. With simulation results, it is shown that the cache-coherent infrastructure outperforms the infrastructure of Woods [1] with a speedup of 1.10. The thesis explores the various configurations of the cache interconnection network and the benefit of the cache-to-cache cache line data transfer with its impact on main memory access. Finally, the thesis shows the cache-coherent infrastructure has very little overhead when using its cache coherence implementation.

Field-Programmable Gate Arrays (FPGAs) systems now comprise many processing elements that are processors running software and hardware engines used to accelerate specific functions. To make the programming of such a system simpler, it is easiest to think of a shared-memory environment, much like in current multi-core processor systems. This thesis introduces a novel, shared-memory, cache-coherent infrastructure for heterogeneous systems implemented on FPGAs that can then form the basis of a shared-memory programming model for heterogeneous systems. With simulation results, it is shown that the cache-coherent infrastructure outperforms the infrastructure of Woods [1] with a speedup of 1.10. The thesis explores the various configurations of the cache interconnection network and the benefit of the cache-to-cache cache line data transfer with its impact on main memory access. Finally, the thesis shows the cache-coherent infrastructure has very little overhead when using its cache coherence implementation.

Distributed generation (DG) is electricity generation by small generating units, which are interconnected at distribution level with capacity less than 50 MW. Environmental concerns and the idea of using cheap and domestic renewable resources increased the popularity of DG following the developments in equipment technology. In Turkey, interconnection of DG is realized through the distribution busbars of 154/36 kV substation. The interconnection of DG at 36 kV feeders is not allowed by distribution system authority. This thesis proposes an interconnection rule which includes technical analyses to be conducted before the permission of interconnection of DG at 36 kV feeders. Moreover, the protection functions and operational requirements needed for the proper and safe operation of distribution system in presence of DG are introduced. A sample distribution system with relevant parameters is used for the simulation studies in Digsilent software. In order to determine the operational reserve requirement against the variations in wind generation, a statistical method including Weibull distribution, standard deviation and monthly average wind speeds is used. Convenience of hydropower plants&rsquo / response for being backup generation against the fluctuations in wind generation is analyzed by a mid-term dynamic model of the power system. A secondary control mechanism for the integration of wind power is suggested. Finally, an economic comparison between the interconnection alternatives of hydropower and photovoltaic power plants at the distribution busbar of the 154/36 kV substation and the 36 kV feeder is done by present worth analysis using the up to date power plant costs and incentives.

As redes-em-chip (NoCs) surgiram como uma alternativa aos atuais problemas de interconexão decorrentes da redução da escala de tecnologia de fabricação de circuitos integrados. O desenvolvimento de transistores com nanômetros de largura tem permitido a inserção de sistemas altamente complexos em uma única pastilha de silício. Dessa forma, os SoCs (Systems-on-Chip) passaram a constituir inúmeros elementos de processamentos (EPs) e as NoCs têm se apresentado como uma opção eficiente no provimento da interconexão dos mesmos, permitindo maior escalabilidade e paralelismo ao sistema. No entanto, esta conexão não é realizada de forma direta. Todo sistema conectado por uma NoC necessita de interfaces de rede (NIs) para intermediar a conexão dos elementos de processamento aos roteadores da rede. O objetivo desse trabalho é apresentar soluções arquiteturais de interfaces de rede para NoCs que atendam diferentes aplicações de forma genérica. Neste trabalho foram desenvolvidas interfaces de redes reutilizáveis e parametrizáveis, e para atender a estas características, as interfaces de rede possibilitam a configuração de diversos parâmetros arquiteturais, como largura da palavra de dados dos EPs, profundidade das FIFOs das interfaces, profundidade das FIFOs da NoC e largura de dados da rede, possibilitando prover a interconexão de qualquer aplicação com um mínimo de reprojeto. As interfaces de rede, juntamente com a NoC, são responsáveis pelo desempenho da comunicação da aplicação e, para isso, o projeto de uma NI deve ser capaz de atender aos requisitos do sistema, por isso, a importância de se obter um projeto de NIs flexível. Para validar as arquiteturas das NIs desenvolvidas, os módulos do decodificador de vídeo no contexto do padrão H.264 foram conectados à NoC através das interfaces projetadas. A partir dessa implementação, puderam-se levantar diversas necessidades que devem ser atendidas pelas NIs. Por fim, foram analisados os resultados de síntese das NIs para diferentes configurações. Também foram verificados os resultados de síntese e desempenho do decodificador de vídeo H.264 conectado pelas NIs à NoC. Com relação aos resultados de síntese em FPGA, a implementação do decodificador de vídeo com NoC e NIs não apresentou um grande aumento em área quando comparada a implementação com conexão ponto-a-ponto. Além disso, para diferentes configurações das NIs, a NoC pode ser utilizada atendendo aos requisitos de desempenho exigidos pela aplicação, sem a necessidade de operar na sua máxima taxa de operação para a resolução QCIF. / Networks-on-Chip (NoCs) have emerged as an alternative to the current interconnection problems arising from the scaling technology for manufacturing integrated circuits. The development of transistors with nanometer-wide has enabled the integration of highly complex systems on a single silicon wafer. Thus, SoCs (Systemson- Chip) have integrated numerous processing elements (EPs) and NoCs have been presented as an effective option in providing the interconnection of these elements, allowing scalability and parallelism to the system. However, this connection is not done directly. Every system connected by NoC needs network interfaces to intermediate the connection of processing elements to network routers. The goal of this thesis is to present architectural solutions for network interfaces for applications in general. In this work we developed a generic, reusable and configurable network interface. The proposed network interface enables the configuration of several architectural parameters, such as data width of the packets, FIFOs depth of the interfaces, FIFO depth and data width of the NoC, and thus, being able to provide the interconnection of any application with a minimal redesign. Network interfaces, together with the NoC, are responsible for application performance and, therefore, the design of an NI should be able to support the system requirements. To validate the architecture of developed NI, the modules of H.264 decoder were connected to NoC through designed interface. From this implementation, one could raise several needs that must be supported by the NIs. Finally, we analyzed the results of synthesis of the NIs for different configurations. It was also analyzed the results of synthesis and performance of H.264 video decoder connected by NIs to NoC. According to results for FPGA synthesis, the implementation of video decoder with NoC and NIs did not show a large increase in area when compared with the implementation of peer-to-peer. Moreover, for different configurations, the NoC can be used according to time requisitions required by the application, without the need to operate at its maximum operation frequency for QCIF resolution.

As redes-em-chip (NoCs) surgiram como uma alternativa aos atuais problemas de interconexão decorrentes da redução da escala de tecnologia de fabricação de circuitos integrados. O desenvolvimento de transistores com nanômetros de largura tem permitido a inserção de sistemas altamente complexos em uma única pastilha de silício. Dessa forma, os SoCs (Systems-on-Chip) passaram a constituir inúmeros elementos de processamentos (EPs) e as NoCs têm se apresentado como uma opção eficiente no provimento da interconexão dos mesmos, permitindo maior escalabilidade e paralelismo ao sistema. No entanto, esta conexão não é realizada de forma direta. Todo sistema conectado por uma NoC necessita de interfaces de rede (NIs) para intermediar a conexão dos elementos de processamento aos roteadores da rede. O objetivo desse trabalho é apresentar soluções arquiteturais de interfaces de rede para NoCs que atendam diferentes aplicações de forma genérica. Neste trabalho foram desenvolvidas interfaces de redes reutilizáveis e parametrizáveis, e para atender a estas características, as interfaces de rede possibilitam a configuração de diversos parâmetros arquiteturais, como largura da palavra de dados dos EPs, profundidade das FIFOs das interfaces, profundidade das FIFOs da NoC e largura de dados da rede, possibilitando prover a interconexão de qualquer aplicação com um mínimo de reprojeto. As interfaces de rede, juntamente com a NoC, são responsáveis pelo desempenho da comunicação da aplicação e, para isso, o projeto de uma NI deve ser capaz de atender aos requisitos do sistema, por isso, a importância de se obter um projeto de NIs flexível. Para validar as arquiteturas das NIs desenvolvidas, os módulos do decodificador de vídeo no contexto do padrão H.264 foram conectados à NoC através das interfaces projetadas. A partir dessa implementação, puderam-se levantar diversas necessidades que devem ser atendidas pelas NIs. Por fim, foram analisados os resultados de síntese das NIs para diferentes configurações. Também foram verificados os resultados de síntese e desempenho do decodificador de vídeo H.264 conectado pelas NIs à NoC. Com relação aos resultados de síntese em FPGA, a implementação do decodificador de vídeo com NoC e NIs não apresentou um grande aumento em área quando comparada a implementação com conexão ponto-a-ponto. Além disso, para diferentes configurações das NIs, a NoC pode ser utilizada atendendo aos requisitos de desempenho exigidos pela aplicação, sem a necessidade de operar na sua máxima taxa de operação para a resolução QCIF. / Networks-on-Chip (NoCs) have emerged as an alternative to the current interconnection problems arising from the scaling technology for manufacturing integrated circuits. The development of transistors with nanometer-wide has enabled the integration of highly complex systems on a single silicon wafer. Thus, SoCs (Systemson- Chip) have integrated numerous processing elements (EPs) and NoCs have been presented as an effective option in providing the interconnection of these elements, allowing scalability and parallelism to the system. However, this connection is not done directly. Every system connected by NoC needs network interfaces to intermediate the connection of processing elements to network routers. The goal of this thesis is to present architectural solutions for network interfaces for applications in general. In this work we developed a generic, reusable and configurable network interface. The proposed network interface enables the configuration of several architectural parameters, such as data width of the packets, FIFOs depth of the interfaces, FIFO depth and data width of the NoC, and thus, being able to provide the interconnection of any application with a minimal redesign. Network interfaces, together with the NoC, are responsible for application performance and, therefore, the design of an NI should be able to support the system requirements. To validate the architecture of developed NI, the modules of H.264 decoder were connected to NoC through designed interface. From this implementation, one could raise several needs that must be supported by the NIs. Finally, we analyzed the results of synthesis of the NIs for different configurations. It was also analyzed the results of synthesis and performance of H.264 video decoder connected by NIs to NoC. According to results for FPGA synthesis, the implementation of video decoder with NoC and NIs did not show a large increase in area when compared with the implementation of peer-to-peer. Moreover, for different configurations, the NoC can be used according to time requisitions required by the application, without the need to operate at its maximum operation frequency for QCIF resolution.

Les dictats de la course à la miniaturisation et à l'accroissement des performances suivit par les industriels de la microélectronique, se heurte aujourd'hui aux limites physiques, technologiques et économiques. Une alternative innovante pour dépasser ces inconvénients, réside en l'intégration tridimensionnelle de circuits intégrés. Cette technologie consiste à empiler verticalement différents niveaux de circuits aux fonctionnalités diverses. Elle ouvre la voie à des systèmes multifonctions ou hétérogènes, aux performances électriques bien meilleures que les circuits bidimensionnels existants. L'empilement de ces puces est réalisable par l'intermédiaire de vias traversant nommés « Though Silicon Via » (« TSV »), qui sont obtenus par la succession de différentes étapes technologiques, dont une d'entre elles consiste à réaliser par gravure plasma, des microcavités profondes à travers le silicium. Actuellement deux procédés de gravure plasma sont principalement utilisés pour la conception de « TSV », le procédé Bosch et le procédé cryogénique, avec dans les deux cas des avantages et des inconvénients différents. L'objet de cette thèse s'inscrit dans le développement d'un procédé de gravure plasma innovant et alternatif à ceux actuellement utilisés, afin de limiter leurs inconvénients (rugosité de flancs, manque de contrôle des profils, basse température…). Dans cette logique deux procédés de gravure profonde ont été envisagés, exploitant les chimies de gravure SF6/O2/HBr et SF6/O2/HBr/SiF4. L'ensemble de l'étude vise à une meilleure compréhension des mécanismes de gravure et de passivation des cavités à fort facteur de forme grâce en particulier à l'exploitation des techniques d'analyse de surface par XPS. / The dictates of miniaturization and increased performance followed by microelectronics manufacturers faces currently physical, technological and economic limitations. An innovative alternative to these problems is the three-dimensional integration of integrated circuits. This technology involves the vertical stacking of different levels of functionality on the various circuits, and thus opens the way for multifunctional or heterogeneous systems, with electrical performance that are much better than those existing in the two-dimensional circuits. The stacking of these chips is achievable through crossing vias named TSV for "Through Silicon Via", which are obtained by the succession of different technological steps,. One of these steps is the realization by plasma etching of deep silicon microcavities. Currently two plasma etch processes are mainly used for the design of TSV or other silicon structures, the Bosch Process and the Cryogenic process, in both cases with different advantages and disadvantages. The purpose of this thesis is to develop an innovative and alternative plasma etching method comparing to those currently used, to minimize their disadvantages (sidewall roughness, lack of profiles control, low temperature ...). In this logic two deep etch processes have been considered, exploiting SF6/O2/HBr and SF6/O2/HBr/SiF4 etching chemistries. All the studies focuses at better understanding of the mechanisms of etching and passivation of high aspect ratio cavities, especially through exploitation of XPS surface analysis

Made available in DSpace on 2015-05-08T14:57:17Z (GMT). No. of bitstreams: 1 arquivototal.pdf: 3191819 bytes, checksum: 142a338f10b5b1c73f589237be4728c2 (MD5) Previous issue date: 2014-02-26 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / This work aims to carry out an analysis of the main existing testing strategies for FPGA, and propose a new strategy applied to the interconnection network of the Xilinx Spartan 3E FPGA based on linear feedback shift register synthesized by Berlekamp Massey Algorithm that can accurately localize the failure. For this, we used softwares from Xilinx manufacturer (specifically, XDL and FPGA_editor) to determine the FPGA based configuration and than create a new proposal and evaluate their employability. As a result of the proposed strategy, it was possible to route 7 WUTs (Wires Under Test) of total of 8 for the FPGA under investigation. Thus, it was necessary 24 test configurations to test and locate the failure on all hexlines and doublelines. The results show that this strategy is able to test 7 WUTs at a time and needs 24 test configurations to test and diagnose precisely the failure location. / Este trabalho objetiva realizar uma análise das principais estratégias de teste já existentes para FPGA, e propor uma nova estratégia aplicada à rede de interconexão do FPGA Xilinx Spartan 3E baseada em registradores de deslocamento com realimentação linear sintetizável pelo Algoritmo de Berlekamp-Massey e que possa diagnósticar com precisão o local da falha. Para isso, foram utilizados softwares da fabricante de FPGAs Xilinx (especificamente, XDL e FPGA_editor) para determinar precisamente a configuração do FPGA e, assim, criar uma nova proposta e avaliar sua empregabilidade. Como resultado, a partir da estratégia adotada foi possível rotear 7 WUTs (Wires Under Test) em um total de 8 para o FPGA em questão. Sendo assim, foram necessárias 24 configurações de teste para testar e diagnósticar todas as linhas do tipo HexLine e DoubleLine. Os resultados obtidos mostram que a estratégia proposta é capaz de testar 7 WUTs por vez e necessita de 24 configurações para testar e diagnósticar precisamente o local da falha na rede de interconexão.

As redes-em-chip (NoCs) surgiram como uma alternativa aos atuais problemas de interconexão decorrentes da redução da escala de tecnologia de fabricação de circuitos integrados. O desenvolvimento de transistores com nanômetros de largura tem permitido a inserção de sistemas altamente complexos em uma única pastilha de silício. Dessa forma, os SoCs (Systems-on-Chip) passaram a constituir inúmeros elementos de processamentos (EPs) e as NoCs têm se apresentado como uma opção eficiente no provimento da interconexão dos mesmos, permitindo maior escalabilidade e paralelismo ao sistema. No entanto, esta conexão não é realizada de forma direta. Todo sistema conectado por uma NoC necessita de interfaces de rede (NIs) para intermediar a conexão dos elementos de processamento aos roteadores da rede. O objetivo desse trabalho é apresentar soluções arquiteturais de interfaces de rede para NoCs que atendam diferentes aplicações de forma genérica. Neste trabalho foram desenvolvidas interfaces de redes reutilizáveis e parametrizáveis, e para atender a estas características, as interfaces de rede possibilitam a configuração de diversos parâmetros arquiteturais, como largura da palavra de dados dos EPs, profundidade das FIFOs das interfaces, profundidade das FIFOs da NoC e largura de dados da rede, possibilitando prover a interconexão de qualquer aplicação com um mínimo de reprojeto. As interfaces de rede, juntamente com a NoC, são responsáveis pelo desempenho da comunicação da aplicação e, para isso, o projeto de uma NI deve ser capaz de atender aos requisitos do sistema, por isso, a importância de se obter um projeto de NIs flexível. Para validar as arquiteturas das NIs desenvolvidas, os módulos do decodificador de vídeo no contexto do padrão H.264 foram conectados à NoC através das interfaces projetadas. A partir dessa implementação, puderam-se levantar diversas necessidades que devem ser atendidas pelas NIs. Por fim, foram analisados os resultados de síntese das NIs para diferentes configurações. Também foram verificados os resultados de síntese e desempenho do decodificador de vídeo H.264 conectado pelas NIs à NoC. Com relação aos resultados de síntese em FPGA, a implementação do decodificador de vídeo com NoC e NIs não apresentou um grande aumento em área quando comparada a implementação com conexão ponto-a-ponto. Além disso, para diferentes configurações das NIs, a NoC pode ser utilizada atendendo aos requisitos de desempenho exigidos pela aplicação, sem a necessidade de operar na sua máxima taxa de operação para a resolução QCIF. / Networks-on-Chip (NoCs) have emerged as an alternative to the current interconnection problems arising from the scaling technology for manufacturing integrated circuits. The development of transistors with nanometer-wide has enabled the integration of highly complex systems on a single silicon wafer. Thus, SoCs (Systemson- Chip) have integrated numerous processing elements (EPs) and NoCs have been presented as an effective option in providing the interconnection of these elements, allowing scalability and parallelism to the system. However, this connection is not done directly. Every system connected by NoC needs network interfaces to intermediate the connection of processing elements to network routers. The goal of this thesis is to present architectural solutions for network interfaces for applications in general. In this work we developed a generic, reusable and configurable network interface. The proposed network interface enables the configuration of several architectural parameters, such as data width of the packets, FIFOs depth of the interfaces, FIFO depth and data width of the NoC, and thus, being able to provide the interconnection of any application with a minimal redesign. Network interfaces, together with the NoC, are responsible for application performance and, therefore, the design of an NI should be able to support the system requirements. To validate the architecture of developed NI, the modules of H.264 decoder were connected to NoC through designed interface. From this implementation, one could raise several needs that must be supported by the NIs. Finally, we analyzed the results of synthesis of the NIs for different configurations. It was also analyzed the results of synthesis and performance of H.264 video decoder connected by NIs to NoC. According to results for FPGA synthesis, the implementation of video decoder with NoC and NIs did not show a large increase in area when compared with the implementation of peer-to-peer. Moreover, for different configurations, the NoC can be used according to time requisitions required by the application, without the need to operate at its maximum operation frequency for QCIF resolution.

[pt] A interligação de mini ou micro-sistemas, seja através de processamento distribuído ou rede, torna-se nos dias atuais muito vantajosa em termos desempenho/custo, quando comparado a sistemas centralizados e via de regra com computadores de grande porte. Este trabalho apresenta um sistema de interligação série orientado para microcomputadores, com controle de barra de comunicação descentralizado. Apresenta como atrativos o baixo custo do hardware necessário à colocação de uma nova estação na rede, o protocolo simples e versátil, modular, e com a possibilidade de colocação de um número ilimitado de estações inteligentes. O seu desempenho pode ser considerado adequado quando utilizado em um ambiente onde não haja excessiva assiduidade à rede; laboratórios universitários e de pesquisa, assim como diversos tipos de controle industrial são bons exemplos. / [en] Mini or micro-systems interconnection, implemented through distributed processing or network, is nowadays more attractive viewed Performance / Cost than main frame centralized huge computers. This work shows a serial network system mainly for usage with microcomputer stations, with descentralized bus control. Main characteristics are low hardware cost to append a new station to the bus, simple protocol, modular design, and ability to support an unlimited number of systems on line. Its performance can be considered adequate when used in environments free from excess of requests to the network bus. University research laboratories, besides several kinds of industrial control applications, may be considered good examples.