Constraint-guided self-adaptation

We present an approach to self-adaptive systems utilizing explicit models of the design-space of the application. The design-space is captured by a hierarchical signal flow representation that allows the specification of alternatives for any component at any level in the model hierarchy. Non-functional requirements and additional knowledge about the system are captured by formal constraints parameterized by operational parameters, such as latency, accuracy, error rate, etc, that are measured at run-time. The constraints and the models are embedded in the running system forming the operation-space of the application. When changes in the monitored parameters trigger a reconfiguration, the operation space is explored utilizing a symbolic constraint satisfaction method relying on Ordered Binary Decision Diagrams. Once a new configuration that satisfies all the constraints is found the reconfiguration takes place.