Software Evolution towards Model-Centric Runtime Adaptivity

Runtime adaptivity is a promising direction towards achieving adaptive behavior for software systems that operate within highly dynamic and non-deterministic environments. Model-centric approaches have proven to be able to successfully address various aspects of runtime adaptivity. In this paper, we propose a target architecture for self-adaptive software systems and show how it facilitates adaptation by interpreting models at runtime. Our approach supports adaptivity using models, which are causally connected to the software application. These models can be queried and transformed dynamically in reaction to changes in the software system's operating environment. We demonstrate how to implement an infrastructure to support the target architecture, and how to prepare and integrate non-adaptive software to comply with this architecture.