Cloud computing is nowadays a popular computing paradigm. Computers are interconnected via network and jointly offer a lot of computing performance. SOFA 2 is a hierarchical component system offering a distributed run-time environment; therefore, it is a suitable environment for cloud computing. Applications are composed from components; each component may run on different computer in the `cloud'. The deployment of the components influences the overall performance of the application and the utilization of resources in the `cloud'; therefore, it has to be planned carefully. In this thesis, an algorithm for automated deployment planning of hierarchical component-based applications is proposed and further implemented in the SOFA 2 system. The algorithm incorporates components' demands and machines' resources in order to maximize performance of the deployed applications. The thesis also proposes and implements extensions that allow using the SOFA 2 component system as an actual cloud platform.

Processes to develop and deliver software have been evolved over the years. One of the primary motivations of this evolution, is gaining the benefits of shorter time-to-market. Continuous deployment is a recent trend to deploy software to the customers automatically and in continuous fashion. Organizations adopting this trend could reach the customers faster through quick deliveries and improve the quality and productivity of the delivered product by an early feedback, and hence achieve increased customer satisfaction. Complex software intensive industrial systems are large-scale, distributed over heterogeneous platforms and interact with several sensors and actuators. Enabling continuous deployment for these industrial systems needs a stable deployment process able to cope with domain specific requirements and challenges. Notably, the required quality attributes of the deployed software product as well as the challenges introduced by the customer-specific nature of the domain. In this thesis, we formalize continuous deployment for industrial systems by identifying the main factors of an appropriate deployment process. In particular, we investigate high-level requirements, required quality attributes of the software product, and challenges in the deployment. Based on this, we propose a continuous deployment pipeline and a set of activities incorporated in the stages of the pipeline, in particular deployment and post-deployment stages. Moreover, we suggest automation support for the activities to both shorten the delivery time and to preserve repeatability and reliability of the deployment process. The aim of such a process is to maintain the quality attributes of the deployed software. We perform a case study to validate the proposed model by implementing a prototype in an industrial system

Les réseaux de capteurs sont utilisés dans l’IoT pour collecter des données. Cependant, une expertise envers les réseaux de capteurs est requise pour interagir avec ces infrastructures. Pour un ingénieur logiciel, cibler de tels systèmes est difficile. Les spécifications des plateformes composant l'infrastructure de capteurs les obligent à travailler à un bas niveau d'abstraction et à utiliser des plateformes hétérogènes. Cette fastidieuse activité peut conduire à un code exploitant de manière non optimisée l’infrastructure. En étant spécifiques à une infrastructure, ces applications ne peuvent également pas être réutilisées facilement vers d’autres infrastructures. De plus, le déploiement de ces applications est hors du champ de compétences d’un ingénieur logiciel car il doit identifier la ou les plateforme(s) requise(s) pour supporter l’application. Enfin, l’architecture peut ne pas être conçue pour supporter l’exécution simultanée d’application, engendrant des déploiements redondants lorsqu’une nouvelle application est identifiée. Dans cette thèse, nous présentons une approche qui supporte (i) la définition de politiques de collecte de données à haut niveau d’abstraction et réutilisables, (ii) leur déploiement sur une infrastructure hétérogène dirigée par des modèles apportés par des experts réseau et (iii) la composition automatique de politiques sur des infrastructures hétérogènes. De ces contributions, un ingénieur peut dès lors manipuler un réseau de capteurs sans en connaitre les détails, en réutilisant des abstractions architecturales disponibles lors de l'expression des politiques, des politiques qui pourront également coexister au sein d'un même réseau. / Sensing infrastructures are classically used in the IoT to collect data. However, a deep knowledge of sensing infrastructures is needed to properly interact with the deployed systems. For software engineers, targeting these systems is tedious. First, the specifies of the platforms composing the infrastructure compel them to work with little abstractions and heterogeneous devices. This can lead to code that badly exploit the network infrastructure. Moreover, by being infrastructure specific, these applications cannot be easily reused across different systems. Secondly, the deployment of an application is outside the domain expertise of a software engineer as she needs to identify the required platform(s) to support her application. Lastly, the sensing infrastructure might not be designed to support the concurrent execution of various applications leading to redundant deployments when a new application is contemplated. In this thesis we present an approach that supports (i) the definition of data collection policies at high level of abstraction with a focus on their reuse, (ii) their deployment over a heterogeneous infrastructure driven by models designed by a network export and (iii) the automatic composition of the policy on top of the heterogeneous sensing infrastructures. Based on these contributions, a software engineer can exploit sensor networks without knowing the associated details, while reusing architectural abstractions available off-the-shelf in their policy. The network will also be shared automatically between the policies.