Section: New Results

Adaptive Middleware

Participants : Aurélien Bourdon, Rémi Druilhe, Damien Fournier, Nicolas Haderer, Gabriel Hermosillo, Jonathan Labejof, Rémi Mélisson, Philippe Merle, Adel Noureddine, Russel Nzekwa, Lionel Seinturier, Daniel Romero, Romain Rouvoy.

In 2011, we pursued our goal to demonstrate that general and high level concepts and solutions can be proposed to design multi-scale middleware systems. The multi-scale aspect has particularly been put forward and we obtained several interesting results: we showed that the concepts of service, component, and software architecture can be successfully used, in the small for wireless sensor middleware platforms [19] , [43] with applications to the Internet of Things (IoT) [27] and for embedded systems [14] , in mid-size distributed environments such as digital home networks [17] , and in the large in cloud computing platforms [30] . We focus below on two achievements which are illustrative in the sense that they address both ends of the targeted spectrum of sizes.

At the scale of small systems, we proposed the REMORA platform [19] , [43] , [27] which defines a lightweight event-based programming model for wireless sensor networks. A C-like language for component implementation and an extension of the state-of-the-art Service Component Architecture (SCA) standard for service-oriented systems are proposed. The platform has been successfully deployed on the Contiki operating system. We showed that despite the characteristics of such resource-constrained environments, we are still able to obtain reconfigurability and adaptability properties for the deployed systems.

At the scale of very large systems, we showed in [30] first results that illustrate the fact that the FraSCAti platform [18] can be used to achieve interoperability between applications deployed on heterogeneous cloud platforms. The experiment is currently deployed on 13 public IaaS and PaaS cloud infrastructures. The very same concepts of service and software architecture that are used at smaller scales are put into practice here. Furthermore, we benefit from the same adaptability properties to address the heterogeneity of concepts needed to fit these very large scale infrastructures.