A Fault-tolerance Framework for Distributed Component Systems

The requirement for higher reliability and availability of systems is continuously increasing even in domains not traditionally strongly concerned by such issues. Required solutions are expected to be efficient, flexible, reusable on rapidly evolving hardware and of course at low cost. Combining both model and component seems to be a very promising cocktail for building solutions to this problem. Hence, we will present in this paper an approach using a model as its first structural citizen all along the development process. Our proposal will be illustrated with an application modeled with UML (extended with some of its dedicated profiles). Our approach includes an underlying execution infrastructure/middleware, providing fault-tolerance services. For the component aspect, our framework promotes firstly an infrastructure based on the Component/Container/Connectorparadigm to provide run-time facilities enabling transparent management of fault-tolerance (mainly fault-detection and redundancy mechanisms). For the model-driven point of view, our framework provides tool support for assisting the users to model their applications and to deploy and configure them on computing platforms. In this paper we focus on the run-time support offered by the component framework, specially the replication-aw are interaction mechanism enabling a transparent replication management mechanisms and some additional system components dedicated to fault-detection and replicas management.