The Internet of Things (IoT) is covering little by little every aspect on our lives. As these systems become more pervasive, the need of managing this complex infrastructure comes with several challenges. Indeed, plenty of small interconnected devices are now providing more than a service in several aspects of our everyday life, which need to be adapted to new contexts without the interruption of such services. However, this new computing system differs from classical Internet systems mainly on the type, physical size and access of the nodes. Thus, typical methods to manage the distributed software layer on large distributed systems as usual cannot be employed on this context. Indeed, this is due to the very different capacities on computing power and network connectivity, which are very constrained for IoT devices. Moreover, the complexity which was before managed by experts on several fields, such as embedded systems and Wireless Sensor Networks (WSN), is now increased by the larger quantity and heterogeneity of the node’s software and hardware. Therefore, we need efficient methods to manage the software layer of these systems, taking into account the very limited resources. This underlying hardware infrastructure raises new challenges in the way we administrate the software layer of these systems. These challenges can be divided into: intra-node, on which we face the limited memory and CPU of IoT nodes, in order to manage the software layer and ; inter-node, on which a new way to distribute the updates is needed, due to the different network topology and cost in energy for battery powered devices. Indeed, the limited computing power and battery life of each node combined with the very distributed nature of these systems, greatly adds complexity to the distributed software layer management. Software reconfiguration of nodes in the Internet of Things is a major concern for various application fields. In particular, distributing the code of updated or new software features to their final node destination in order to adapt it to new requirements, has a huge impact on energy consumption. Most current algorithms for disseminating code over the air (OTA) are meant to disseminate a complete firmware through small chunks and are often implemented at the network layer, thus ignoring all guiding information from the application layer. First contribution: A models@runtime engine able to represent an IoT running application on resource constrained nodes. The transformation of the Kevoree meta-model into C code to meet the specific memory constraints of an IoT device was performed, as well as the proposition of modelling tools to manipulate a model@runtime. Second contribution: Component decoupling of an IoT system as well as an efficient component distribution algorithm. Components decoupling of an application in the context of the IoT facilitates its representation on the model@runtime, while it provides a way to easily change its behaviour by adding/removing components and changing their parameters. In addition, a mechanism to distribute such components using a new algorithm, called Calpulli is proposed.