Runtime dynamic reconfiguration of field-programmable gate arrays (FPGAs) and devices incorporating microprocessors and FPGA has been successfully utilized to increase performance and reduce power consumption. While previous methods have been successful, they typically do not consider the runtime behavior of the application that can be significantly affected by variations in data inputs, user interactions, and environmental conditions. In this dissertation, we present a dynamically reconfigurable system and design methodology that optimizes performance and power consumption by determining which coprocessors to implement with an FPGA based upon the current application behavior.For dynamically reconfigurable systems, in which the selection of hardware coprocessors to implement within the FPGA is determined at runtime, online estimation methods are essential to evaluate the performance and power consumption impact of the hardware coprocessor selection. We present a base profile assisted online system-level performance and power estimation framework for estimating the speedup and power consumption of dynamically reconfigurable embedded systems.Importantly though, complex interactions between multiple application tasks, non-deterministic execution behavior, and effects of operating system scheduling introduce significant challenges. To address these, we further present an adaptive online performance and power estimation framework suing kernel speedup coefficient adaptation that monitors and adapts the changing application and system behavior for multitasked applications. By exhaustively examining predefined voltage and frequency settings for the microprocessor and hardware kernels, the potential speedup and power reduction can be effectively estimated for each configuration and voltage/frequency settings. These estimates can be utilized to determine the optimal system configuration. At the same time, the kernel speedup coefficients for each kernel can be dynamically updated to account for the difference between the estimated and actual performance measured at runtime.Finally, in order to quickly determine kernel selection and voltage and frequency settlings, we present an efficient, online heuristic performance and power estimation framework that significantly decreases execution time at the cost of a small increase in power consumption. This online heuristic estimation framework achieves significant power reduction compared to software only implementation without performance degradation.

Thèse diffusée initialement dans le cadre d'un projet pilote des Presses de l'Université de Montréal/Centre d'édition numérique UdeM (1997-2008) avec l'autorisation de l'auteur.

The project suggests an alternative means for living in Qatar. It proposes the development of a transient, floating community a man-made, transitory archipelago of floating housing units located at the ‘soon to be abandoned’ docks a short distance from the Museum of Islamic Art and the Doha Corniche. The design, through a variety of bespoke dwelling options, can provide both more and less nomadic housing to accommodate a variety of dif- ferent lifestyles and social units. Clustered around three permanent islands (containing a cultural/activity center, three adaptable ‘work-unit’ towers, and a park/commercial area), the design provides a wide set of adaptive layout patterns within which the mobile units can be arrayed. Directly plugged into the city, the community functions as a floating appendix to downtown Doha: an adaptable and flexible city-supplement that can expand or contract ac- cording to need, whether to aid in the accommodation of guests for various large-scale sporting or other events, or merely to provide an alternative locale for long or short stay visitors to Doha alike.

The history of housing in Sweden is a central part of our cultural history and for the collective understanding of ourselves. During the 20th century the Swedish state has played an active roll in the production of housing. But, since the 90s the housing question has been left to the market which struggles to produce the quantity of housing that is needed. While the housing that is built today comes with a lot of qualities, it is also ridden with problems. The floor plans are general, yet fixed, common spaces are under prioritized and the possibility to customize your apartment is low. Not to mention that we are moving towards a situation were also rental apartments are put on the free market which under the current situation, could lead to an increased segregation in the bigger cities. The ability to choose how your home should be configurated has become a luxury.   The situation has certain similarities with the speculative way of building of the late 19th century in Sweden. The answer then was to make it possible for workers, small farmers and officials to build their own homes with financial help from the state.   When neither the market nor the state can provide good enough housing for the citizens, it is time to, like before, explore how we can take the matter in our own hands. Today an answer could be joint building ventures. Friends, colleagues or like-minded can get together and build their own multifamily houses by planning, building and finally inhabit the buildings. It is today a tricky process, but experiences from Germany have showed that it can be a reliable way of providing housing when the idea gets more normative. Holmsund could be a good testing ground for joint building ventures. The settlement is expected to grow, there are free central plots and the settlement can provide enough service and commuting possibilities even for people more used to city life.   My system provides a frame, concrete slabs resting on steel columns with a wet core stabilizing the structure laterally. The users are then, in dialogue with the architect and the other residents negotiating the space after their needs and economical situation.

[pt] Este trabalho propõe um gráfico de controle EWMA para observações individuais ou médias amostrais, com a constante de amortecimento variando entre dois valores de acordo com o valor mais recente da estatística EWMA, para obter detecção mais rápida de alterações pequenas a moderadas na média do processo, e sem a complexidade operacional apresentada por outros esquemas adaptativos, pois o tamanho da amostra e o intervalo de amostragem são mantidos fixos. Já existe um outro trabalho propondo a variação da constante de amortecimento dos gráficos EWMA, mas com base em outro critério: Capizzi e Masarotto (2003). O esquema EWMA adaptativo foi combinado com limites de Shewhart para os valores individuais (ou médias amostrais), para acelerar a detecção de grandes deslocamentos da média do processo, também sem aumento da complexidade operacional. Os NMA1´s - números esperados de amostras até um sinal verdadeiro - foram calculados por um método de aproximação numérica usando um modelo matemático por cadeias de Markov, e comparados com os do esquema EWMA tradicional (com parâmetros fixos) e com os do esquema adaptativo de Capizzi e Masarotto (2003). O esquema proposto tende a fornecer NMA1´s menores para alterações na média acima de 1,0 desvio-padrão, e o esquema de Capizzi e Masarotto (2003) tende a fornecer NMA1´s menores para pequenas alterações. Ambos os esquemas possuem melhor desempenho que o gráfico EWMA com parâmetros fixos. Uma vantagem que pode se tornar decisiva para a adoção do esquema proposto é a simplicidade dos cálculos requeridos para o monitoramento. / [en] This work proposes an EWMA process control chart for individual observations or subgroup averages, in which the smoothing constant varies between two values according to the most recent value of the EWMA statistic, in order to achieve faster detection of small to moderate shifts in the process mean, and without the operational complexities presented by other adaptive schemes, since its sample size and sampling interval do not vary. There is one other work proposing the adaptive variation of the smoothing constant of EWMA charts, but based on a different criterion: Capizzi and Masarotto (2003). The adaptive EWMA scheme was combined with Shewhart limits for the individual values (or subgroup averages), to enhance its sensitivity to large shifts, again with no extra operational burden. The out-of-control average run lengths (ARL1´s) were calculated through a numerical approximation method based on a Markov chain model. The ARL1´s were compared of the proposed scheme, of the traditional (fixed parameter) EWMA chart and of Capizzi and Masarottos´s adaptive EWMA scheme. The proposed scheme generally provides the shortest ARL1´s for shifts in the mean above one standard deviation, and Capizzi and Masarotto´s scheme tends to outperform it for smaller shifts. Both schemes perform better than the fixed parameter EWMA. An advantage that can become decisive for the adoption of the proposed scheme is the simplicity of the calculations required for the monitoring.

abstract: Infrastructure are increasingly being recognized as too rigid to quickly adapt to a changing climate and a non-stationary future. This rigidness poses risks to and impacts on infrastructure service delivery and public welfare. Adaptivity in infrastructure is critical for managing uncertainties to continue providing services, yet little is known about how infrastructure can be made more agile and flexible towards improved adaptive capacity. A literature review identified approximately fifty examples of novel infrastructure and technologies which support adaptivity through one or more of ten theoretical competencies of adaptive infrastructure. From these examples emerged several infrastructure forms and possible strategies for adaptivity, including smart technologies, combined centralized/decentralized organizational structures, and renewable electricity generation. With institutional and cultural support, such novel structures and systems have the potential to transform infrastructure provision and management. / Dissertation/Thesis / Masters Thesis Civil, Environmental and Sustainable Engineering 2018

This research proposed the use of state-of-the-art wireless communications and networked embedded systems technologies to provide environmental sensing for the early detection of natural disasters. The data is acquired, processed and transmitted, from the location where the disaster originates, to potentially threatened conurbations in order to promptly notify the population. The acquired data is transformed from its raw form into information that can be utilized by local authorities to rapidly assess emergency situations and then to apply disaster management procedures. Alternatively, the system can generate alerting signals without human intervention. Furthermore, recorded historical data can be made available for scientists to build models, to understand and to forecast the behavior of disasterous events. An additional, important, contribution of this research was the analysis and application of Wireless Sensor Network technology for disaster monitoring and alerting.

Cette thèse présente une approche pour la construction d'évaluateurs de requêtes adaptés aux besoins des applications et/ou à l'environnement d'exécution. L'approche proposée distingue trois types d'adaptation : statique (à la construction de l'évaluateur), personnalisée (avant l'exécution de requête ayant des contraintes spécifiques) et dynamique (pendant l'exécution de la requête). Nous avons présenté une analyse systématique des différentes dimensions d'un évaluateur. La principale contribution de cette thèse est un canevas d'évaluation de requêtes nommé QBF (Query Broker Framework). QBF présente les fonctionnalités d'un évaluateur de requêtes de manière abstraite et séparée dans le but de favoriser leur réutilisation et leur adaptation. En s'appuyant sur QBF, nous avons proposé les mécanismes pour assurer les trois types d'adaptation, à savoir l'instanciation de QBF, l'évaluation personnalisée et l'évaluation interactive de requêtes. Une implémentation de QBF ainsi que de quelques instances (i.e. évaluateurs) ont été réalisées. Elles nous ont permis de montrer la faisabilité de notre approche et le bénéfice de la séparation et de l'abstraction des fonctionnalités proposées dans QBF. Nous avons également mené une évaluation expérimentale des mécanismes implémentés (i.e. opérateurs, surveillances) afin de mesurer le surcoût de l'adaptation dans QBF et d'aider les programmeurs à choisir des éléments (ou composants) appropriés pour construire leurs évaluateurs.

Notre objectif dans cette thèse est de donner la propriété d'adaptabilité à l'aspect duplication. La séparation des préoccupations et l'approche par services permettent au développeur d'applications de s'abstraire de l'aspect duplication lors de ses développements. Cependant, ces approches souffrent d'une limitation majeure : il semble très difficile, voir impossible, de fournir un service/aspect générique de duplication pouvant être paramétré afin d'être utilisé dans différents contextes d'exécution et couvrant l'ensemble des protocoles existants. Ces constatations nous ont conduit à la définition d'un canevas de services de duplication, nommé RS2.7. RS2.7 est le squelette d'un service de duplication définissant sa structure. Il permet d'obtenir des services de duplication indépendants de tout code propre à l'application, pouvant être utilisés dans différents contextes non fonctionnels (transactionnel, mémoires partagées, etc.) et prenant en compte les contraintes et les protocoles spécifiques à chaque domaine. Nos contributions portent sur trois axes : (1) la modélisation des services de duplication pouvant être obtenus à partir de RS2.7, (2) l'adaptabilité du canevas par rapport au contexte non fonctionnel et (3) l'adaptabilité dans tout ou partie des protocoles de duplication. RS2.7 a été mis en \oe uvre et notre validation porte sur la démonstration des caractéristiques d'adaptabilité offertes. Nous cherchons à montrer que notre canevas permet d'obtenir des services très variés et convenant pour divers contextes non fonctionnels.

Les administrateurs, les développeurs, les concepteurs logiciels ont besoin de nouvelles approches pour faire face à la complexité croissante des systèmes informatiques. Afin de libérer les administrateurs de tâches répétitives et d'améliorer la réactivité des systèmes, ces systèmes, alors appelés systèmes autonomiques, tendent à être dotés de capacités d'auto-gestion telles que l'auto-configuration et l'auto-réparation. Ces capacités d'auto-gestion sont fortement couplées à la capacité du système à s'auto-adapter, c.-à-d. à modifier sa structure et son comportement alors qu'il est en cours d'exécution. Les approches se sont jusqu'alors concentrées sur l'auto-adaptation structurelle des systèmes à base de composants logiciels en remaniant les liaisons entre les composants. Dans notre approche, nous nous focalisons sur l'adaptation comportementale. Nous avons défini le modèle de composants MOCAS (Model Of Components for Adaptive Systems) pour permettre la réalisation de systèmes autonomiques à base de composants logiciels auto-adaptatifs. MOCAS repose sur une approche uniforme exploitant l'ingénierie des modèles : les capacités d'adaptation et les propriétés autonomiques de chaque composant reposent sur l'usage qui est fait par MOCAS du langage de modélisation UML. La structure d'un composant MOCAS repose sur les éléments natifs du langage MOCAS. Le comportement du composant est décrit avec une machine à états MOCAS. Cette machine est ensuite exécutée par le composant pour réaliser son comportement. Pour devenir adaptable, le composant est déployé dans un conteneur respectant ce même modèle. Une boucle de contrôle à base de composants MOCAS permet de réaliser des composants auto-adaptables de façon décentralisée et de les doter de propriétés autonomiques. Les politiques autonomiques exploitent alors les modèles de machines à états afin d'auto-configurer et d'auto-réparer les composants.