Un certain nombre de personnes m’ont demandé ce que je pouvais bien aller faire au Japon. Vu que ces temps, j’ai surtout été préoccupé par les démarches administratives et les préparatifs, cet aspect de la question m’est un peu sorti de la tête. Je pars 18 mois au Japanese Advanced Institute of Science and Technology. Là bas, je travaillerais sur un projet de recherche concernant les communications de groupe, payé par le Fond National Suisse pour la recherche.
Voici le résumé en anglais du projet :
Group communication is a set of abstractions to handle a set of distributed processes as a single entity. Group communication primitives typically offer precise and strong properties even in the event of link and process failures. Group communication is meant to simplify the design of distributed applications, in particular fault-tolerant applications. Despite more than twenty years of research in the domain of group communication and tremendous progress in the domain of networking, group communication has seen little use outside of academic prototypes.
As group communication toolkits are typically implemented assuming different system models and offer primitives with complex specifications, building an application using group communication is a non-trivial task. Yet the design of applications based on group communication has not been studied extensively; in fact, there is a lack of literature on the subject. Finally, as there are many different and incompatible toolkits, an application can only be targeted to a single toolkit.
We believe that the lack of acceptance of group communication comes from the fact that group communication systems have typically been designed in isolation, with little consideration for the existing network infrastructure or application level issues.
The goal of our project is to address these issues by designing a group communication toolkit that integrates with the overall network infrastructure, and offers group communication primitives suited for real applications, primitives using standard interfaces.
The project will have two main contributions. First, it will present how different applications can use group communication. Second, it will show how a toolkit can be built using standard low level services, offering group communication functionalities encapsulated in standard interfaces. The design will be validated by a prototype implementation.