Le développement orienté-service et le cloud computing offrent beaucoup d'opportunités au développement et au déploiement d'applications. En effet, le développement orienté-service permet de composer des fonctionnalités issues de services distribués, développés par différentes organisations. D'un autre côté, le cloud computing permet de provisionner des environnements évolutifs (en fonction du besoin en ressources) de développement et de déploiement, à la demande. Dans ce travail de recherche, nous proposons et décrivons une méthode de provisonnement automatique d'applications métier, orientées-service sur le cloud. Nous avons appelé cette méthode MADONA (Method for AutomateD prOvisioning of service-oriented busiNess Applications). MADONA couvre le cycle de vie de provisionnement d'applications et est basée sur un orchestrateur de services pour la gestion de la configuration, du déploiement, et de la composition de services métier.Dans ce travail, nous visons à réduire les connaissances techniques nécessaires au provisionnement d'applications métier. Pour ce faire, nous apportons trois contributions majeures. Premièrement, l'automatisation de ce provisionnement. En effet, les phases de MADONA sont complètement automatisées. L'utilisateur n'intervient que pour exprimer son besoin et pour utiliser l'application métier automatiquement générée (par la composition de services métier) et déployée sur une IaaS présélectionnée. Deuxièmement, l'enrichissement de la description des services par des concepts liés aux relations d'un service. En effet, les langages de description de services décrivent le plus souvent ces derniers comme des entités isolées et ne considèrent pas les relations entre services. Nous avons défini dans ce travail les relations de composition qui décrivent pour chaque service métier les services nécessaires à son bon fonctionnement, et les services avec lesquels il peut être composé.Troisièmement, nous permettons à l'utilisateur d'exprimer son besoin à un haut niveau d'abstraction des détails techniques de composition et de déploiement. Nous avons pour cela défini un vocabulaire pour formaliser ces besoins fonctionnels (en termes de mots clés décrivant les fonctionnalités désirées, ou de noms de services désirés) et non fonctionnels (en termes de coût, de préférences de déploiement (ex : localisation de déploiement), et de qualité de service). La méthode a été prototypée et testée suivant plusieurs scénarios montrant sa faisabilité / Service-oriented computing and cloud computing offer many opportunities for developing and deploying applications. In fact, service-oriented computing allows to compose several functionalities from distributed services developed by different organizations. On the other hand, cloud computing allows to provision on demand scalable development and deployment environments. In this resarch work, we propose and describe a Method for AutomateD prOvisioning of service-oriented cloud busiNess Applications (MADONA). The method covers the whole application’s lifecycle and is based on cloud orchestration tools that manage the deployment and dependencies of supplied components. This research work aims to reduce the necessary technical knowledge for provisioning service-oriented cloud applications. To this end, we bring three major contributions. Firstly, we automatise the whole application provisioning. In fact, MADONA phases are fully automated. The user intervenes only in requirement elicitation and when the application is deployed and ready to use. The business application is automatically generated (by composing business services) and deployed in an automatically preselected IaaS. Secondly, we enrich the description of services by integrating concepts describing services’ interactions. In fact, service description languages usually describe services as isolated components and does not consider the interactions between services. We define in this work, composition interactions which describe for each business service, its necessary services and the services with which it can be composed. Thirdly, we allow the user to express her requirements abstracting composition and deployment technical details. To this end, we defined a RequIrement VocAbuLary (RIVAL) to formalize user’s functional (in terms of keywords describing the desired functionalities, or names of desired services) and non-functional requirements (in terms of cost, deployment preferences (ex. preferred location), and quality of service). The method has been implemented and tested showing its faisability

O Paradigma Orientado a Objetos tem sido atualmente a abordagem dominante de desenvolvimento de software. Contudo, ela sofre da Tirania da Decomposição Dominante, pois não permite uma modularização adequada da implementação relativa a interesses estruturais. Como consequência, a implementação relativa a cada interesse estrutural fica espalhada pelos módulos do programa e entrelaçada com a implementação relativa a outros interesses estruturais. Outras abordagens de desenvolvimento de software, como o Desenvolvimento de Software Orientado a Aspectos com AspectJ e a Separação Multidimensional de Interesses em Hiperespaços com Hyper/J e CME, atingem sucesso moderado em oferecer mecanismos que permitem superar as deficiências do Paradigma Orientado a Objetos. No entanto, tais abordagens também possuem deficiências e omissões que devem ser reparadas para que elas possam se tornar utilizáveis em contextos típicos de desenvolvimento de software complexo. Este trabalho especifica uma nova abordagem, denominada Desenvolvimento de Software Orientado a Temas (DSOT), que tem como objetivo superar algumas deficiências das abordagens anteriores por meio de mecanismos que permitem a manipulação da implementação de cada interesse estrutural de forma separada e a manipulação da implementação de cada tipo de dado de forma separada. Além disso, DSOT possui operadores que são ortogonais, isto é, podem ser utilizados de forma combinada ou separada, para efetuar a composição de módulos do programa. Mostra-se o modelo conceitual do DSOT e descrevese um estudo de caso que consiste no desenvolvimento de um programa para demonstrar mais concretamente como o DSOT funciona na prática. Não se demonstra a superioridade do DSOT para o caso geral, mas os resultados alcançados evidenciam que o DSOT é uma abordagem promissora que merece ser investigada mais aprofundadamente em pesquisas futuras / The Object-Oriented Paradigm has currently been the dominant approach for developing software. However, it suffers from the Tyranny of the Dominant Decomposition, as it does not support a suitable modularization to the implementation relative to structural concerns. As a consequence, the implementation relative to each structural concern is scattered throughout the program modules and tangled with the implementation relative to other structural concerns. Some software development approaches, such as Aspect-Oriented Software Development with Aspect and Multidimensional Separation of Concerns in Hyperspaces with Hyper/J and CME, achieve moderate success in offering mechanisms that make it possible to overcome the deficiencies of the Object-Oriented Paradigm. However, such approaches also possess deficiencies and ommissions that must be corrected in order for them to get usable in typical complex software development contexts. This work specifies a new approach, named Theme- Oriented Software Development (TOSD), which aims at overcoming some deficiencies from previous approaches through mechanisms that support the handling of implementation for every structural concern separately and the handling of implementation for every data type separately. Moreover, TOSD contains operators which are orthogonal, that is, they can be used separately or as a combination, in order to perform composition of the program modules. We show the conceptual model of TOSD and describe a case study which consists in the development of a program to demonstrate more concretely how TOSD works in practice. We do not demonstrate the superiority of TOSD for the general case, but the results we have obtained suggest that TOSD is a promissing approach which deserves a deeper investigation in future research

O Paradigma Orientado a Objetos tem sido atualmente a abordagem dominante de desenvolvimento de software. Contudo, ela sofre da Tirania da Decomposição Dominante, pois não permite uma modularização adequada da implementação relativa a interesses estruturais. Como consequência, a implementação relativa a cada interesse estrutural fica espalhada pelos módulos do programa e entrelaçada com a implementação relativa a outros interesses estruturais. Outras abordagens de desenvolvimento de software, como o Desenvolvimento de Software Orientado a Aspectos com AspectJ e a Separação Multidimensional de Interesses em Hiperespaços com Hyper/J e CME, atingem sucesso moderado em oferecer mecanismos que permitem superar as deficiências do Paradigma Orientado a Objetos. No entanto, tais abordagens também possuem deficiências e omissões que devem ser reparadas para que elas possam se tornar utilizáveis em contextos típicos de desenvolvimento de software complexo. Este trabalho especifica uma nova abordagem, denominada Desenvolvimento de Software Orientado a Temas (DSOT), que tem como objetivo superar algumas deficiências das abordagens anteriores por meio de mecanismos que permitem a manipulação da implementação de cada interesse estrutural de forma separada e a manipulação da implementação de cada tipo de dado de forma separada. Além disso, DSOT possui operadores que são ortogonais, isto é, podem ser utilizados de forma combinada ou separada, para efetuar a composição de módulos do programa. Mostra-se o modelo conceitual do DSOT e descrevese um estudo de caso que consiste no desenvolvimento de um programa para demonstrar mais concretamente como o DSOT funciona na prática. Não se demonstra a superioridade do DSOT para o caso geral, mas os resultados alcançados evidenciam que o DSOT é uma abordagem promissora que merece ser investigada mais aprofundadamente em pesquisas futuras / The Object-Oriented Paradigm has currently been the dominant approach for developing software. However, it suffers from the Tyranny of the Dominant Decomposition, as it does not support a suitable modularization to the implementation relative to structural concerns. As a consequence, the implementation relative to each structural concern is scattered throughout the program modules and tangled with the implementation relative to other structural concerns. Some software development approaches, such as Aspect-Oriented Software Development with Aspect and Multidimensional Separation of Concerns in Hyperspaces with Hyper/J and CME, achieve moderate success in offering mechanisms that make it possible to overcome the deficiencies of the Object-Oriented Paradigm. However, such approaches also possess deficiencies and ommissions that must be corrected in order for them to get usable in typical complex software development contexts. This work specifies a new approach, named Theme- Oriented Software Development (TOSD), which aims at overcoming some deficiencies from previous approaches through mechanisms that support the handling of implementation for every structural concern separately and the handling of implementation for every data type separately. Moreover, TOSD contains operators which are orthogonal, that is, they can be used separately or as a combination, in order to perform composition of the program modules. We show the conceptual model of TOSD and describe a case study which consists in the development of a program to demonstrate more concretely how TOSD works in practice. We do not demonstrate the superiority of TOSD for the general case, but the results we have obtained suggest that TOSD is a promissing approach which deserves a deeper investigation in future research