A self-adaptation strategy for service-based architectures

Self-adaptive software systems are known to respond at run time to changes detected internally or in their environment, in an attempt to keep meeting their own functional requirements and agreed levels of service. Such response usually targets their architectures and involve, in particular, the possibility of their dynamic reconfiguration. In contexts where change is the rule rather than the exception, it is difficult to predict when exactly such reconfigurations are needed, and if they will lead the system into a suitable configuration. However, knowing the main attributes of the context, it is possible to plan configurations that will be more likely to perform well in some conjugation of values for such attributes. In this paper we discuss both a model that lays down reconfiguration strategies, planned at design time, and a strategy which actively uses such a model to trigger architectural adaptations at run time. This strategy builds on a framework intended to the formal verification of architectural requirements, either from a qualitative or quantitative (probabilistic) perspective.