How to manage a dynamic environment and how to provide task partitioning are two key concerns when developing distributed computing applications. The emergence of Grid computing environments extends these problems. Conventional resource management systems are based on a relatively static resource model and a centralized scheduler that assigns computing resources to users. Distributed management introduces resource heterogeneity: not only the set of available resources, but even the set of resource types is constantly changing. Obviously this is unsuitable for the present Grid. In addition, the Grid provides users with the physical infrastructure to run parallel programs. Because of this increasing availability, there are more requirements for parallelization technologies. Therefore, based on problems outlined above, this thesis provides a novel scheduler which not only enables dynamic management but also provides skeleton library to support the task partition. Dynamic management is derived from the concept of reflectiveness, which allows the Grid to perform like an efficient market with some limited government controls. To supplement the reflective mechanism, this thesis integrates a statistical forecasting approach to predict the environment of the Grid in the next period. The task partitioning support is extended from the skeleton library in the parallel computing and cluster computing areas. The thesis shows how this idea can be applied in the Grid environment to simplify the user’s programming works. Later in this PhD thesis, a Petri-net based simulation methodology is introduced to examine the performance of the reflective scheduler. Moreover, a real testing environment is set up by using a reflective scheduler to run a geometry optimization application. In summary, by combining knowledge from economics, statistics, mathematics and computer science, this newly invented scheduler not only provides a convenient and efficient way to parallelize users’ tasks, but also significantly improves the performance of the Grid.

How to manage a dynamic environment and how to provide task partitioning are two key concerns when developing distributed computing applications. The emergence of Grid computing environments extends these problems. Conventional resource management systems are based on a relatively static resource model and a centralized scheduler that assigns computing resources to users. Distributed management introduces resource heterogeneity: not only the set of available resources, but even the set of resource types is constantly changing. Obviously this is unsuitable for the present Grid. In addition, the Grid provides users with the physical infrastructure to run parallel programs. Because of this increasing availability, there are more requirements for parallelization technologies. Therefore, based on problems outlined above, this thesis provides a novel scheduler which not only enables dynamic management but also provides skeleton library to support the task partition. Dynamic management is derived from the concept of reflectiveness, which allows the Grid to perform like an efficient market with some limited government controls. To supplement the reflective mechanism, this thesis integrates a statistical forecasting approach to predict the environment of the Grid in the next period. The task partitioning support is extended from the skeleton library in the parallel computing and cluster computing areas. The thesis shows how this idea can be applied in the Grid environment to simplify the user’s programming works. Later in this PhD thesis, a Petri-net based simulation methodology is introduced to examine the performance of the reflective scheduler. Moreover, a real testing environment is set up by using a reflective scheduler to run a geometry optimization application. In summary, by combining knowledge from economics, statistics, mathematics and computer science, this newly invented scheduler not only provides a convenient and efficient way to parallelize users’ tasks, but also significantly improves the performance of the Grid.

L'augmentation du nombre des sinistres à grande échelle d'origine climatique, technologique, sanitaire, et bien d'autres, nous interpelle sur le mode de gouvernance des systèmes concernés, sur les choix de couverture de ces risques et sur les mesures de prévention mises en place. La question qui a animé notre recherche est : dans quelle mesure la gouvernance d'un système favorise-t-elle la gestion des risques ? Notre analyse empirique s'appuie sur deux cas d'étude, qui se concentrent, sur la gouvernance d'un système face à des risques catastrophiques pour l'un, et des risques communs pour l'autre. D'une part, les risques majeurs sont connus comme étant très peu probables mais avec de possibles lourdes conséquences. La dynamique de la gouvernance adaptative, pour la gestion des risques majeurs d'un système, souligne la nécessité d'adopter une approche flexible (dynamique). Nous proposons ainsi un cadre d'évaluation de la capacité d'adaptation d'un système dans le domaine des risques majeurs. D'autre part, les risques communs affectent l'ensemble de la population. Face aux risques communs dont les conséquences s'étalent dans le futur, nous considérons que la gouvernance adaptative dans une démocratie délibérative se réfère au paradigme selon lequel la gouvernance permet aux décideurs de se projeter vers le long terme. Le dilemme consiste à prendre (à présent) des décisions (c.-à-d. des risques, au sens général du terme) dont les effets (ou les conséquences) ne se font sentir que dans le futur. La gouvernance d'un système doit favoriser la gestion des risques. Notre travail de recherche est structuré autour de trois parties (i.e. « Gouvernances et risques », « La gouvernance d'un système exposé à un risque catastrophique », et « La gouvernance d'un système exposé à un risque commun »), chacune composée de trois chapitres. L'objectif ultime de la gestion des risques, voire des incertitudes qui caractérisent nos sociétés contemporaines est une raison d'être légitime des modes dynamiques de gouvernance qui incitent les différents acteurs à travailler ensemble, de diverses manières, afin de relever les défis auxquels ils sont confrontés. / The increase in large-scale climate, technological, sanitary, and many other types of losses, challenges us on the governance of the concerned systems, the choices made to hedge these risks and implement prevention measures. The question which animated our research is : to what extent the governance of a system promotes a better risk management?Our empirical analysis is based on two case studies which focus on the governance of a system while facing catastrophic risks for first one, and on common risks for the other. On the one hand, catastrophic risks are known to be highly improbable but with possible serious consequences. The dynamic of the adaptive governance system facing catastrophic risks underlines the need for a (dynamic) flexible approach. We therefore suggest an assessment framework for the adaptive capacity of a system in the field of catastrophic risks. On the other hand, common risks affect the whole population. We consider that the adaptive governance in a deliberative democracy refers to the governance paradigm which addresses common risks whose consequences spread out into the future. Thus, it induces decision makers to look at the long run. The dilemma consists in taking (now) decisions (i.e. Risk, in the general sense of the word) whose effects (or consequences) are only felt in the future. The governance of a system shall promote risk management. Our research is structured into three parts (i.e. "Governance and risk", "Governance of a system exposed to a catastrophic risk" and "Governance of a system exposed to a common risk"), each composed of three chapters. The ultimate objective of the management of risks, or even uncertainties, which characterize our contemporary societies is a legitimate reason for having dynamic modes of governance that prompt different actors to work together in various ways, in order to meet challenges they face.

The survival and sustainable development of construction contractors depends on their own capacity as well as their ability to cope with the ever changing environment. Hong Kong was once a large construction market by the world standards; but in the period 1998 to 2007 has suffered from long term market shrinkage. This shrinkage in the work load available in the market has demanded that Hong Kong contractors change their business strategy and consider expanding their construction business to the international marketplace. From surveys and studies of Hong Kong contractors, it is noted that traditionally the majority of contractors organizations were not active in participating in the international construction markets. The prime reasons were that the Hong Kong domestic market was large enough to sustain the contractors and that Hong Kong construction professionals are reluctant to work overseas. However due to the continuous shrinkage of the market, Hong Kong contractors were being forced to consider expanding their construction business to the overseas markets in order to survive. This thesis reviews existing theories and previous studies in overseas con-struction business development. It analyzes surveys of Hong Kong con-struction contractors and conducts a scientific study of a Hong Kong based contractor organization which expanded its construction business to overseas markets. This study was conducted through an integrated action research methodology. Based on the problems, impacts, difficulties and success this organization faced during its practical experience in the overseas business expansion, this thesis explores and recommends a structured approach for Hong Kong contractors to re-define and develop their business overseas. The focus of study is the pre-contract award stage of business development. It focuses on the entry modes, strategic planning, risk management and tender management of the organization. The author also presents various models for use in attaining width and depth of understanding of overseas knowledge. These include preliminary entry selection model, dynamic management, and a spiral model (a learning & knowledge based business development model), for use in the management of international construction business development at the pre-contract stage. The groundwork laid down in this thesis will form the basis for further studies and the development of theories / models. It is the expectation of the author that other contractors in Hong Kong and other nation s contractors may make use of this research as assistance to their overseas business development. The ultimate aim of the author has been to change the goal of contractors from being companies able to compete with international contractors in their domestic market to contractors that are able to compete in the international market place.

Αντικείμενο της παρούσας διατριβής υπήρξε αφενός η μελέτη και η αξιολόγηση των υπαρχόντων τεχνικών και αφετέρου ο σχεδιασμός, η υλοποίηση και η αποτίμηση νέων τεχνικών και μηχανικών για τη βελτιστοποίηση της απόδοσης του συστήματος με τη βέλτιστη αξιοποίηση των διαθέσιμων πόρων. / The objective of this dissertation was on the one hand the study and analysis of existing techniques and algorithms for large-scale DVEs and on the other hand the design, implementation and assessment of new techniques, methods and algorithms that can ensure the quality of the provided services and the efficient operation of the system. The algorithms and techniques developed were analysed and evaluated through extensive experiments both for identifying possible weaknesses and for highlighting the applications that can achieve optimum results. More specifically, two approaches have been designed and developed. The first approach is related to the design and development of a partitioning algorithm, that, in contrary to existing approaches, exploits the special characteristic of the virtual world and the impact those characteristics have on users’ behavior in order to predict the needs of the environment, in terms of resources, before they are required. The second approach is related to the design and development of a dynamic management approach that exploits the dynamic nature of DVEs in order to perform load balancing and to ensure performance optimization. During the elaboration of this dissertation, we found out that one of the basic problems when designing and implementing algorithms, methods and techniques for large-scale DVEs is the way that their efficiency could be examined. In most of the cases, the evaluation is based on theoretical models, which often fail to meet the circumstances and situations met in real DVEs. In particular, for supporting largescale DVEs, extended infrastructure is needed in terms of both hardware and software. Due to the fact that both researchers and application designers do not always have access to such extended infrastructure, the assessment and evaluation of developed techniques are extremely difficult. In most of the cases, both application designers and researchers adopt specialized methods for evaluating different techniques, while in other cases simulation tools have been developed from scratch. However, given the fact that the design and implementation is application or technique-specific, the reusability of these tools for different architectures and algorithms is not always successful. To this direction and for overcoming this important limitation, a simulation modeling framework for assessing DVEs performance was designed and implemented. The framework takes into account a number of both generic and special-cause parameters, which can be set on demand by the DVE designers and stake-holders and it is based on transforming system requirements to the concepts of operational management.

L’émergence de l’internet des objets (IoT), les applications de types réseaux de capteurs et les nouveaux réseaux pour les objets nécessite le développement d’un nouvel écosystème. De nos jours, elle bouscule plusieurs secteurs de notre société. Cela, sollicite la conception des émetteurs-récepteurs radios fréquence à basse consommation étant donné que la réduction de la consommation d’énergie présente une contrainte importante dans le cas de ces applications afin d’obtenir une autonomie accrue. Dans ce contexte l’objet des travaux de thèse est de proposer des techniques de réduction de la consommation des récepteurs radio fréquence tout en cherchant à minimiser l’impact de ces techniques sur les performances des récepteurs réalisés. Dans l’optique de réaliser un démonstrateur composé d’un émetteur récepteur permettant une transmission vidéo, deux récepteurs UWB impulsionnel non cohérent à gestion dynamique d’énergie ont été réalisés en technologie HCMOS9 0.13µm de STMicroelectronics. Dans un premier temps, une étude des techniques de gestion dynamique d’alimentation sur les circuits analogiques radio fréquences a été proposée. Cette étude a été menée sur différents circuits qui semblent être le plus utilisés en conception de circuits analogiques à hautes fréquences. La technique proposée permet d’allumer et éteindre les circuits entre deux impulsions reçues afin de réduire leur consommation. L’application de cette technique nécessite par ailleurs une réduction du temps de latence causé par l’allumage et les extinctions des fonctions radio fréquence. Dans ce cas, un modèle permettant de minimiser l’impact de l’effet d’encapsulation a été proposé. / The emergence of the Internet of Things (IoT), sensors networks and new networks to objects requires the development of a new ecosystem. Nowadays, it upsets many sectors of our society. It solicits design of low power radio transceivers as reducing energy consumption presents a major constraint in the case of these applications in order to obtain greater autonomy. In this context, the purpose of the thesis is to provide techniques allowing reducing the power consumption of radio frequency receivers while seeking to minimize the impact of these technologies on the performance of the achieved receiver. In order to realize a demonstrator consists of a transmitter and receiver for video transmission, two UWB receivers with dynamic power management have been made in 0.13µm HCMOS9 technology from STMicroelectronics. First, a study of dynamic power management techniques on analog radio frequency circuits was proposed. This study was conducted on different circuits that seem to be the most used in design of analog circuits at high frequencies. The proposed technique allows to turn on and off the circuit between two pulses received to reduce their consumption. The application of this technique also requires a reduction of the latency caused by the ignition and the extinction radio frequency functions. In this case, a model to minimize the impact of the encapsulating effect has been proposed. Secondly, the first receiver was performed for 6-10GHz frequency band and implements dynamic power management using the technique of "Power Gating".

Aujourd'hui, alors que le nombre de marqueurs génétiques disponibles augmente rapidement, de nouvelles populations doivent être créées pour exploiter au mieux cette quantité d'informations dans le but de mieux comprendre l'architecture génétique de caractères complexes. Les populations de type MAGIC ont été créées pour rassembler les avantages des populations bi-parentales et des panels d'associations, la bonne puissance de détection et une localisation précise. L'objectif de cette thèse était d'étudier l'intérêt de la population MAGIC INRA pour l'analyse de l'architecture génétique de la précocité de floraison. Cette population a été créée à partir de 60 parents brassés durant 12 générations de panmixie grâce à l'introduction d'un gène de stérilité mâle (ms1b). Cette étude a été réalisée sur 56 parents toujours disponibles en banque de graines et 380 lignées dérivées de la population après les 12 générations de recombinaison. Cette population a été génotypée avec la puce 9K iSelect, représentant environ 5 000 SNPs localisés sur tout le génome, additionnée de 14 marqueurs localisés dans des gènes candidats. Ce jeu de données moléculaires a été complété par des données fines de phénotypage de la précocité de floraison. Suite aux 12 générations de panmixie, le DL de cette population a été très réduit, à longue comme à moyenne distance (<10cM). Ce faible DL nous a amené à développer un algorithme basé uniquement sur le DL qui ordonne les marqueurs de manière à avoir un DL décroissant monotone avec la distance. L'algorithme ordonne globalement de la même manière que la carte génétique les marqueurs à longue distance mais à courte distance le DL est moins lié à la distance génétique. La différence réside sur l'équilibre entre les effets de la recombinaison et de la dérive génétique sur le DL. L'intérêt de la population MAGIC INRA pour détecter des QTLs a ensuite été étudié avec deux approches : une approche évolutionniste et une approche de génétique d'association. La première approche détecte les loci soumis à sélection par comparaison des fréquences alléliques de la population initiale (G0) et de la population évoluée (G12) grâce à une nouvelle méthode. La population initiale est composée des parents pondérés par une contribution estimée avec une nouvelle méthode bayésienne. 26 régions génomiques soumises à sélection ont été détectées. Une analyse de génétique d'association avec les marqueurs détectés sous sélection a montré que respectivement cinq et trois zones étaient associées à la précocité avec un semis d'automne et au caractère printemps/hiver. Une analyse phénotypique a effectivement mis en évidence la précocification de la date de floraison et une augmentation de la proportion de plantes de type printemps. Une analyse de génétique d'association a ensuite été réalisée sur les lignées SSD sur 12 caractères x environnements i.e. la date d'épiaison et le temps de remplissage du grain mesurés dans six environnements. Les tests d'association ont aussi été réalisés avec des variables synthétisant l'information présente dans plusieurs traits phénotypiques soit avec une ACP, soit avec un modèle écophysiologique. Au total, toutes ces analyses ont détecté six QTLs dont trois correspondants à des gènes majeurs. Parmi ces six QTLs, deux sont spécifiques des caractères mesurés avec un semis d'automne et deux avec ceux mesurés avec un semis de printemps. / Nowadays, with the dramatically increase of available molecular markers, there is a deep need for new populations allowing to exploit all of this information to better understand the genetic architecture of complex traits. MAGIC populations as they are built to bring together bi-parental populations and association panel advantages, provide such powerful detection and fine mapping capacities. The aim of these PhD was to study the MAGIC INRA population usefulness for the study of genetic architecture of earliness. This population is derived from 12 cycles of random crosses between 60 founders, turning wheat from selfing to outcrossing thanks to the use of a nuclear male sterility gene (ms1b, Probus donor). This population is composed of 56 parents still available and 380 SSD lines. Parents and SSD lines were genotyped using the 9K iSelect SNPs array, providing around 5 000 SNPs on the whole genome, as well as 14 addition markers located in candidate genes. They were also finely phenotyped for earliness traits. With the 12 panmictic generations, the population LD decreased strongly, especially at long and medium distance (<10cM). This allowed us to develop an algorithm mapping markers on the sole pairwise LD information, ordering markers in a way to have the LD decreasing along the distance. When considering long distances, overall the results were consistent with the order found on genetic maps while at short distance LD was poorly linked to genetic distance. These differences between long and short distances were linked to the balance between recombination and drift effects on LD. The usefulness of the MAGIC INRA population for QTL detection was analyzed with two approaches: an evolutionary approach and an association genetics approach. The first one detects loci under selection by identifying high shift in allelic frequency with a new method. The initial population was composed of founders weighted by a contribution estimated with a new Bayesian method. 26 genomic areas under selection were detected. An association genetics analysis with the markers detected as under selection showed respectively five and three genomic regions associated with earliness and growth habit. Actually the G12 population was found phenotypically earlier than the G0 and with more spring individuals. A broader association genetics analysis was performed on G12 population, studying 12 traits x environments i.e. heading date and grain filling time, both observed in six environmental conditions. Two additional integrated traits from either PCA or ecophysiological model were also analyzed. In all, these different analyses detected six QTLs, three of them corresponding to candidate genes. Among these six QTLs, two were specific to autumn sowing and two specific to spring sowing.

Made available in DSpace on 2014-07-29T14:57:51Z (GMT). No. of bitstreams: 1 dissertacao-rodrigohs.pdf: 1383844 bytes, checksum: b3fda2012ce5a20dc390677f308520e3 (MD5) Previous issue date: 2012-05-29 / Context models that describe dynamic context-aware scenarios usually need to be frequently updated. Some examples of situations that motivate these updates are the appearance of new services and context providers, the mobility of the entities described in these models, among others. Generally, updates on models imply redevelopment of the architectural components of context-aware systems based on these models. However, as these updates in dynamic scenarios tend to be more frequent, it is desirable that they occur at runtime. This dissertation presents an infrastructure for dynamic management of context models based on the fundamentals of complex event processing, or CEP. This infrastructure allows the fundamental abstractions from which a model is built to be updated at runtime. As these updates can impact systems based on the updated models, this dissertation identifies and analyzes these impacts, which are reproduced in a case study that aims to evaluate the proposed infrastructure by demonstrating how it deals with the impacts mentioned. / Modelos contextuais que descrevem cenários de computação sensível ao contexto dinâmicos normalmente precisam ser frequentemente atualizados. Alguns exemplos de situações que motivam essas atualizações são o surgimento de novos serviços e provedores de informações contextuais, a mobilidade das entidades descritas nesses modelos, dentre outros. Normalmente, atualizações em modelos implicam em redesenvolvimento dos componentes arquiteturais dos sistemas sensíveis ao contexto baseados nesses modelos. Porém, como em cenários dinâmicos essas atualizações tendem a ser mais frequentes, é desejável que elas ocorram em tempo de execução. Essa dissertação apresenta uma infraestrutura para gerenciamento dinâmico de modelos de contexto baseada nos fundamentos de processamento complexo de eventos, ou CEP. Essa infraestrutura permite que as abstrações fundamentais a partir das quais um modelo é construído sejam atualizadas em tempo de execução. Como essas atualizações podem causar impactos nos sistemas baseados nos modelos atualizados, essa dissertação identifica e analisa esses impactos, os quais são reproduzidos em um estudo de caso que tem como finalidade avaliar a infraestrutura proposta através da demonstração de como ela lida com os impactos mencionados.