O desenvolvimento de programas paralelos e distribuídos encontra na programação baseada em passagem de mensagens uma abordagem eficaz para explorar adequadamente as características das máquinas de memória distribuída. Com o uso de clusters e de bibliotecas de suporte às trocas de mensagens, como o padrão MPI (Message Passing Interface), aplicações eficientes e economicamente viáveis podem ser construídas. Em tais sistemas, o tempo despendido nas comunicações constitui um importante fator de desempenho a ser considerado e requer a utilização de procedimentos e cuidados para a sua correta caracterização. Neste trabalho, modelos analíticos de primitivas de comunicação bloqueante MPI são desenvolvidos segundo uma metodologia de análise e predição apropriada. São tratadas algumas das principais operações ponto-a-ponto e coletivas e, utilizando técnicas de ajuste de curvas e tempos experimentais, o comportamento das primitivas de comunicação é representado em equações, possibilitando ainda a realização de análises e predições de desempenho em função do tamanho das mensagens e do número de processos envolvidos. Através de testes em um cluster de estações de trabalho, a precisão dos modelos elaborados é comprovada. Sendo a maioria dos erros percentuais inferiores a 8%, os resultados obtidos confirmam a validade do processo de modelagem. Além disso, o trabalho apresenta um conjunto de funções construídas com o objetivo de oferecer suporte a atividades de análise e predição, procurando facilitar e automatizar sua execução. / The development of parallel and distributed programs finds at message-passing programming a powerful approach to explore properly the distributed memory machines issues. Using clusters and message-passing libraries, as MPI standard (Message Passing Interface), efficient and cost effective applications can be constructed. In these systems, the time spent with communications means a important performance factor to be considered and its correct characterization requires procedures and cautions. In this work, analytic models for MPI blocking communication primitives are developed according one appropriate methodology for analysis and prediction. Some of the main peer-to-peer and collective operations are treated, and through curve fitting techniques and experimental times the behavior of the communication primitives is represented in equations, allowing also the accomplishment of performance analysis and prediction in function of the message length and the number of processes. Tests realized in a cluster of workstations prove the accuracy of the elaborated models. With most of errors within 8%, the obtained results show the validity of the modeling process. Also, the work presents a set of functions constructed with the purpose of support analysis and prediction activities, in order to facilitate and automate them.

Existem muitos fatores capazes de influenciar o desempenho de um programa paralelo MPI (Message Passing Interface). Dentre esses fatores, podemos citar a quantidade de dados processados, o número de nós envolvidos na solução do problema, as características da rede de interconexão, o tipo de switch utilizado, entre outros. Por isso, realizar predições de desempenho sobre programas paralelos que utilizam passagem de mensagem não é uma tarefa trivial. Com o intuito de modelar e predizer o comportamento dos programas citados anteriormente, nosso trabalho foi desenvolvido baseado em uma metodologia de análise e predição de desempenho de programas paralelos MPI. Inicialmente, propomos um modelo gráfico, denominado DP*Graph+, para representar o código das aplicações. Em seguida, desenvolvemos modelos analíticos, utilizando técnicas de ajuste de curvas, para representar o comportamento das estruturas de repetição compostas por primitivas de comunicação e/ou computação local. Além disso, elaboramos modelos para predizer o comportamento de aplicações do tipo mestre/escravo. Durante o desenvolvimento das atividades de análise e predição de desempenho, implementamos algumas funções para automatizar tarefas e facilitar nosso trabalho. Por último, modelamos e estimamos o desempenho de duas versões diferentes de um programa de multiplicação de matrizes, a fim de validar os modelos propostos. Os resultados das predições realizadas sobre os programas de multiplicação de matrizes foram satisfatórios. Na maioria dos casos preditos, os erros ficaram abaixo de 6 %, confirmando a validade e a precisão dos modelos elaborados. / There are many factors able to influence the performance of a MPI (Message Passing Interface) parallel program. Within these factors, we may cite: amount of data, number of nodes, characteristics of the network and type of switch, among others. Then, performance prediction isn’t a easy task. The work was developed based on a methodology of analysis and performance prediction of MPI parallel programs. First of all, we proposed a graphical model, named DP*Graph+, to represent the code of applications. Next, we developed analytical models applying curve fitting techniques to represent the behavior of repetition structure compounds by comunication primitives and/or local computations. Besides, we elaborated models to predict aplications of type master/slave. For development of performance prediction activities, some functions was developed to automate tasks and make our work easy. Finally, we modeled and predicted the performance of two different programs of matrix multiplication to prove the accuracy of models. The results of predictions on the programs were good. In the majority of predicted cases, the errors were down 6 %. With these results, we proved the accuracy of developed models.

No mundo globalizado em que vivemos os sistemas, as aplicações e as pessoas precisam estar conectadas permanentemente à Internet, e a uma variedade de redes de comunicação, com vários dispositivos, simultaneamente. O ideal, devido a esse contexto, seria ter um único dispositivo com uma única interface de rede e um único programa que permitisse fazer diversas conexões com diversos protocolos usados simultaneamente, tornando o sistema simples de usar e fácil de instalar e operar, conseguindo, desta forma, o nível de estabilidade e confiabilidade desejadas. Por isso, as redes de área pessoal (WPAN) estão crescendo em importância desde seu aparecimento em 1991. Diversas tecnologias foram propostas para sua implementação sem que haja um consenso sobre a aplicabilidade de cada opção. Este trabalho realiza uma avaliação crítica das diversas tecnologias propostas e por meio de um laboratório de prototipação procurou obter dados que permitam uma avaliação mais objetiva dessas tecnologias. Esse documento apresenta uma descrição desse laboratório e apresenta a metodologia proposta para comparação das tecnologias para WPAN, e os resultados obtidos em laboratório. O estudo apresentado demonstra, através de uma avaliação critica, que as principais alternativas no momento são Wi-Fi e Bluetooth e que com os dados obtidos na prototipação em laboratório se provou que hoje há possibilidade de virtualizar diferentes redes e tipos de interface em uma única interface física. A economia de escala, multiplicada pelo efeito da virtualização das interfaces, leva a conclusão que as tecnologias Wi-Fi e Bluetooth deverão dominar o mercado de WLAN e WPAN nos próximos anos. / In todays globalized world, systems, applications and people need to be permanently connected to the Internet, a variety of communications networks and several different devices simultaneously. Ideally, faced with this context, there should be a single device with a single network interface, and a single program that enables several connections and protocols to be used simultaneously, making the entire systems simple to use and easy to install and operate, thus leading to the desired levels of stability and reliability. So the personal area networks (WPAN) are growing in importance since its debut in 1991. Several technologies have been proposed for its implementation without a consensus on the applicability of each option The proposed work will conduct a critical evaluation of the various proposed technologies and, through a laboratory prototype seek information to enable a more objective evaluation of these technologies. This document provides a description of the laboratory and presents the proposed methodology to compare technology for WPAN.

Esta tese apresenta um estudo sobre a implementação de sistema de detecção e reconhecimento de faces no processador CELL na plataforma CBE, utilizando um sistema Playstation 3. Inicialmente, diversas abordagens para reconhecimento e detecção de faces são estudadas, bem como arquiteturas de processador multi núcleos. São apresentadas três implementação, sendo a segunda implementação premiada com quarto colocado no IBM CELL UNIVERSITY CHALLENGE 2007 para desenvolvimento de programas para plataforma Cell BE. A terceira implementação apresenta os resultados interessantes relacionados a vetorização do processamento dos dados da detecção de objetos e os recursos adotados para obter o melhor desempenho. / This thesis presents a study of face detection implementation on CBE plataform and employ the system with Playstation 3 hardware. Several approaches for face detection and recognition are studied as well as multicore processor architetures. We implemented three versions of system. First implementation was a naive reference implementation with worst performance. Second implementation granted fourth prize in IBM CELL UNIVERSITY CHALLENGE 2007 that incentive development on CBE plataform. Third implementation had most interesting results with vectorized approaches on code of object detection.

Esta tese parte da hipótese de que nem todos os edifícios de escritório com certificação ambiental na cidade de São Paulo, construídos entre os anos 2000 a 2016, atendem de modo satisfatório as condições de iluminação natural, segundo as recomendações hoje em vigor. Esses edifícios contemporâneos possuem grandes áreas envidraçadas em suas fachadas, sem proteção solar, que causam problemas de conforto térmico e luminoso. Para evitar o ofuscamento, os usuários acionam proteções solares internas como telas e persianas, o que acaba por impedir ou diminuir o acesso à luz natural. A pesquisa realizada fez um levantamento dos padrões construtivos prevalecentes em 28 edifícios de escritório que se enquadram na especificação descrita acima para verificar questões como: a profundidade da planta, transmissão luminosa dos vidros, WWR (Wall Window Ratio), entre outras. Os dados encontrados serviram para a elaboração de modelos referenciais, usados em simulações computacionais, com o objetivo de verificar a influência de variáveis arquitetônicas no desempenho da iluminação natural. As simulações foram realizadas no plug-in Diva do programa Rhinoceros 3D, que simula a iluminação natural de forma integrada com os softwares Radiance/DAYSIM, já validados pela comunidade científica internacional. O tratamento dos dados foi feito através de um método inédito de avaliação do desempenho luminoso e os resultados obtidos mostraram que 60% dos cenários analisados foram reprovados. Esta tese propõe uma metodologia de avaliação do desempenho luminoso de edifícios de escritório que considere de modo integrado a disponibilidade de luz natural, o conforto visual e o consumo de energia. / This thesis is based on the hypothesis that not all office buildings with environmental certification in the city of São Paulo, constructed between 2000 and 2016, meet satisfactorily the conditions of daylighting, according to the recommendations currently in force. These contemporary offices have large glass areas without sun protection in their facades, causing problems of thermal and luminous comfort. In order to avoid glare, users employ indoor shading devices such as rollers and blinds, which ultimately prevent or reduce access to daylight. This research has carried out a survey on the prevailing constructive patterns of 28 office buildings that follow the specifications described above, to verify issues such as plan depth, light transmission of glasses, WWR (Wall Window Ratio), among others. The collected data were used to elaborate reference models used in computational simulations, aiming at verifying the influence of architectural variables on the performance of daylighting. The simulations were carried out with the Diva plug-in for Rhinoceros 3D program, which simulates daylighting, integrated with software Radiance/DAYSIM, already validated by the international scientific community. The data treatment was carried out through an innovative method for evaluating luminous performance, and the obtained results showed that 60% of the analyzed scenarios were disapproved. This thesis proposes that the methodology for evaluating the luminous performance of office buildings should consider the availability of daylight, visual comfort, and energy consumption in an integrated way.

Le modèle de programmation à base d'acteurs a été intensivement utilisé pour le développement de grandes applications et systèmes. On citera par exemple la fonction chat de Facebook ou bien encore WhatsApp. Ces systèmes peuvent avoir plusieurs milliers d'utilisateurs connectés simultanément avec des contraintes fortes de performance et d'interactivité. Ces systèmes s"appuient sur des infrastructures informatiques basées sur des processeurs multi-cœurs. Ces infrastructures disposent en général d'un espace mémoire partagé et hiérarchique NUMA (Non-Uniform Memory Access). Notre analyse de l'état de l'art montre que peu d'études ont été menées sur l'adéquation des environnements d'exécution à base d'acteurs avec des plates-formes à mémoire hiérarchique. Ces environnements d'exécution font en général l'hypothèse que l'espace de mémoire est complètement plat, ce qui pose ensuite de sérieux problèmes de performance. Dans cette thèse, nous étudions les défis posés par les plates-formes multi-cœurs à mémoire hiérarchiques pour des environnements à base d'acteurs. Nous étudions plus particulièrement les problèmes de gestion mémoire, d'ordonnancement et d'équilibrage de charge.Dans la première partie de la thèse, nous avons analysé et caractérisé les applications basées sur le modèle d'acteurs. Cela a permis de mettre en évidence le fait que les exécutions des applications et benchmarks faisaient ressortir des structures de communication particulières que les environnements d'exécution se doivent de prendre en compte pour optimiser les performances. La prise en compte du graphe de communication et la mise en œuvre ont été effectuées dans un environnement d'exécution réel, la machine virtuelle (VM) du langage de programmation Erlang. Le langage de programmation Erlang s'appuie sur le modèle d'acteurs avec une syntaxe claire et cohérente pour la gestion des acteurs. Les modifications que nous avons intégrées à la machine virtuelle Erlang permettent d'améliorer significativement les performances grâce à une meilleure prise en compte de l'affinité entre des acteurs qui interagissent beaucoup. L'ordonnancement et la régulation de charge de l'application sont également améliorées grâce à une meilleure connaissance de l'application et de la topologie de la plate-forme. Une des perspectives serait d'intégrer ces contributions à d'autres environnements d'exécution à base d'acteurs, comme par exemple ceux des Kilim et Akka. / The actor model is present in several mission-critical systems, such as those supporting WhatsApp and Facebook Chat. These systems serve thousands of clients simultaneously, therefore demanding substantial computing resources usually provided by multi-processor and multi-core platforms. Non-Uniform Memory Access (NUMA) architectures account for an important share of these platforms. Yet, research on the the suitability of the current actor runtime environments for these machines is very limited. Current runtime environments, in general, assume a flat memory space, thus not performing as well as they could. In this thesis we study the challenges hierarchical shared memory multi-core platforms present to actor runtime environments. In particular, we investigate aspects related to memory management, scheduling, and load-balancing.In this document, we analyze and characterize actor based applications to, in light of the above, propose improvements to actor runtime environments. This analysis highlighted the existence of peculiar communication structures. We argue that the comprehension of these structures and the knowledge about the underlying hardware architecture can be used in tandem to improve application performance. As a proof of concept, we implemented our proposal using a real actor runtime environment, the Erlang Virtual Machine (VM). Concurrency in Erlang is based on the actor model and the language has a consistent syntax for actor handling. Our modifications to the Erlang VM significantly improved the performance of some applications thanks to better informed decisions on scheduling and on load-balancing. As future work we envision the integration of our approach into other actor runtime environments such as Kilim and Akka.

Les techniques de programmations pour le calcul de haute performanceont adopté les modèles basés sur parallélisme de tâche qui sontcapables de s’adapter plus facilement à des superordinateurs avec desarchitectures hybrides. La performance des applications basées surtâches dépende fortement des heuristiques d'ordonnancement dynamiqueset de sa capacité à exploiter les ressources de calcul et decommunication.Malheureusement, les stratégies d'analyse de performancetraditionnelles ne sont pas convenables pour comprendre les supportsd'exécution dynamiques et les applications basées sur tâches. Cesstratégies prévoient un comportement régulier avec des phases decalcul et de communication, par contre, des applications basées surtâches ne manifestent pas de phases précises. Par ailleurs, la granularitéplus fine des applications basées sur tâches typiquement provoque descomportements stochastiques qui donnent lieu aux structuresirrégulières qui sont difficiles à analyser.Dans cette thèse, nous proposons des stratégies d'analyse deperformance qui exploitent la combinaison de la structure del'application, d'ordonnancement et des informations de laplate-forme. Nous présentons comment nos stratégies peuvent aider àcomprendre des problèmes de performance dans des applications baséesur tâches qui exécutent dans des plates-formes hybrides. Nosstratégies d'analyse de performance sont construites avec des outilsmodernes pour l'analyse de données, ce que permettre la création despanneaux de visualisation personnalisés. Ces panneaux permettent lacompréhension et l'identification de problèmes de performancesoccasionnés par de mauvaises décisions d'ordonnancement etconfiguration incorrect du support d'exécution et de laplate-forme. Grâce à combinaison de simulation et débogage nouspouvons aussi construire une représentation visuelle de l'état interneet des estimations calculées par l'ordonnancer durant l'ordonnancementd'une nouvelle tâche.Nous validons notre proposition parmi de l'analyse de tracesd'exécutions d'une factorisation de Cholesky implémenté avec lesupport d'exécution StarPU et exécutée dans une plate-forme hybride(CPU/GPU). Nos études de cas montrent comment améliorer la partitiondes tâches entre le multi-(GPU, coeur) pour s'approcher des bornesinférieures théoriques, comment améliorer le pipeline des opérationsMPI entre le multi-(noeud, coeur, GPU) pour réduire le démarrage lentedans les noeuds distribués et comment optimiser le support d'exécutionpour augmenter la bande passante MPI. Avec l'emploi des stratégies desimulation et débogage, nous fournissons un workflow pourl'examiner, en détail, les décisions d'ordonnancement. Cela permet deproposer des changements pour améliorer les mécanismes d'ordonnancementet prefetch du support d'exécution. / Programming paradigms in High-Performance Computing have been shiftingtoward task-based models that are capable of adapting readily toheterogeneous and scalable supercomputers. The performance oftask-based applications heavily depends on the runtime schedulingheuristics and on its ability to exploit computing and communicationresources.Unfortunately, the traditional performance analysis strategies areunfit to fully understand task-based runtime systems and applications:they expect a regular behavior with communication and computationphases, while task-based applications demonstrate no clearphases. Moreover, the finer granularity of task-based applicationstypically induces a stochastic behavior that leads to irregularstructures that are difficult to analyze.In this thesis, we propose performance analysis strategies thatexploit the combination of application structure, scheduler, andhardware information. We show how our strategies can help tounderstand performance issues of task-based applications running onhybrid platforms. Our performance analysis strategies are built on topof modern data analysis tools, enabling the creation of customvisualization panels that allow understanding and pinpointingperformance problems incurred by bad scheduling decisions andincorrect runtime system and platform configuration.By combining simulation and debugging we are also able to build a visualrepresentation of the internal state and the estimations computed bythe scheduler when scheduling a new task.We validate our proposal by analyzing traces from a Choleskydecomposition implemented with the StarPU task-based runtime systemand running on hybrid (CPU/GPU) platforms. Our case studies show howto enhance the task partitioning among the multi-(GPU, core) to getcloser to theoretical lower bounds, how to improve MPI pipelining inmulti-(node, core, GPU) to reduce the slow start in distributed nodesand how to upgrade the runtime system to increase MPI bandwidth. Byemploying simulation and debugging strategies, we also provide aworkflow to investigate, in depth, assumptions concerning the schedulerdecisions. This allows us to suggest changes to improve the runtimesystem scheduling and prefetch mechanisms.

Considerando-se o contexto de gerenciamento energético em grades móveis, neste trabalho foram propostos dois algoritmos de escalonamento (Maximum Regret e Greedy) que, além de minimizar o consumo de energia, visam assegurar o cumprimento dos requisitos de qualidade de serviço das aplicações submetidas pelos usuários. Tais algoritmos foram projetados a partir de soluções heurísticas para o problema de escalonamento ciente de consumo de energia em grades móveis, que foi modelado como um problema de otimização envolvendo variáveis binárias. Por meio de experimentos, que consideraram tanto cenários estáticos quanto dinâmicos, foi demonstrada a viabilidade dos algoritmos de escalonamento propostos em relação à redução do consumo de energia. Em seu pior caso, o algoritmo Maximum Regret foi 12,18% pior que o referencial determinado pela melhor solução do solver Gurobi; já no pior caso do algoritmo Greedy, tal diferença foi de apenas 8,14% / Considering the context of energy management in mobile grids, this work proposes two scheduling algorithms (Maximum Regret and Greedy) that aim not only to reduce the energy consumption of the mobile devices, but also to ensure the QoS (Quality of Service) requirements of the running applications. These algorithms were designed based on heuristics for the energy aware scheduling problem in mobile grids, which was modeled as an optimization problem with integer variables. The performances of the proposed scheduling algorithms were evaluated by an extensive set of experiments, which demonstrated the feasibility of the adopted approach regarding energy consumption minimization. In its worst case, the Maximum Regret algorithm was 12.18% worse than the best solution provided by the Gurobi solver. While in the Greedys worst case the performance difference was just 8.14%

U.S. Coast Guard leaders have received feedback concerning gaps in performance management of the Marine Inspection Program (MIP) from maritime industry stakeholders, Department of Homeland Security representatives, and internal agents over the past decade. The purpose of this case study was to explore strategies to improve performance in the U.S. Coast Guard MIP. Data were gathered through a review of documentation pertinent to marine inspection (i.e., policy, requirements, analyses, reports, and job aids) and 13 semistructured interviews with personnel from 3 distinct organizational levels. Study participants represented civilian and active duty personnel from all geographical U.S. Coast Guard districts, as well as tactical, strategic, and policy levels of the MIP. The conceptual framework of the study was Fusch and Gillespie's human competence model. Data analysis was based on coding of words, phrases, and sentences from multiple sources of data to identify recurring themes through methodological triangulation. The thematic analysis of the study data revealed themes that included lack of mission clarity, limited information management resources, differences in skills and knowledge management among inspectors, and unclear requirements for selecting a marine inspector. The study framework provided a basis for additional performance management research in government entities. The recommendations from this study may lead to social change through improved U.S. Coast Guard marine inspection services, which could result in greater safety, reduced pollution, and fewer security risks in the navigable waterways of the United States.

A methodology was developed to conduct diagnostic performance assessment and simulation of alternative delivery scenarios in pressurized irrigation distribution networks. It consists of three components, an agro-hydrologic model able to forecast peak water demand hydrographs, a hydraulic model with capability of simulating the network behavior under different flow configurations, and a set of performance indicators for conducting assessments of performance achievements relative to specified targets. As a preliminary work, the current delivery schedule of an existing pressurized irrigation network (system 1) and the resulting effects on crop irrigation management were analyzed by simulating soil water balance and irrigation scheduling at field level. Simulations allowed analyzing the on-farm irrigation management under the current rotation deliveries, and comparing it with an alternative flexible irrigation scheduling to maximize crop yields. Results at field level were up-scaled to the entire command area of the system, showing the usefulness of soil water balance and of irrigation scheduling as analytical tools to demonstrate the inconsistency between the current water delivery and crops’ and farmers’ requirements. This preliminary work also allowed highlighting the need for modernizing the irrigation management in the first of the two study areas considered for the present research work (system 1 located in the province of Taranto, southern Italy). An existing agro-hydrological model conceived to forecast water demand hydrographs in pressurized delivery networks was enhanced through several refinements and amendments of the computation algorithms. The refined model was applied for validation at different management levels on an existing pressurized irrigation system (system 2) located in the province of Foggia, southern Italy, where water withdrawals by farmers and the main hydraulic parameters are recorded on a continuous basis for monitoring purposes. Results from validation showed that the model is capable of forecasting with good accuracy the timing of peak-demand periods, the seasonal demand irrigation volumes, as well as the hydrographs of hourly flow rates demanded by farmers during these peak periods, especially when it is applied to large multi-cropped command areas. Performance indicators, originally conceived for diagnostic assessment in canal systems, were modified for application to pressurized distribution networks, and reference standard values were proposed. These indicators were then applied for validation to the second study area (irrigation system 2), where records of water deliveries are available, and showed their usefulness for diagnostic performance assessments.