Joint Effort of Features and Components to Safely Reconfigure Software Architectures

Adapting software to user needs and environmental constraints at run-time requires extending the software life cycle with a reconfiguration step to go further than the analysis, design, development, and first deployment. Existing approaches for reconfiguring cloud computing and cyber-physical systems involving software compute new valid configurations either at design time, at runtime, or both. To avoid computational or validation overheads that are induced by these approaches for each reconfiguration step, we propose to rely on variability models to acquire a representation of the set of valid configurations of a system. More precisely, architecture defined by feature models allow us to automatically generate component-based models while taking into account some logical or behavioral constraints. Produced run-time variability models in JavaBIP introduce monitoring and controlling the application behavior by dealing with reconfiguration requests while ensuring the validity of all reachable configurations. Experimental results show the interest of our approach in practice by evaluating the computational overhead induced by generated run-time models on system architectures involving up to 500 features