RE-PREF: Support for REassessment of PREFerences of Non-functional Requirements for Better Decision-Making in Self-Adaptive Systems

Modelling and reasoning with prioritization of non-functional requirements (NFRs) is a research field that needs more attention. We demonstrate RE-PREF, an approach that supports the modelling of NFRs and their preferences, and discovery of possible scenarios where badly chosen preferences can either make the runtime system miss or suggest unnecessary adaptations that may degrade the behavior of a self-adaptive system (SAS). Specifically, we showcase how RE-PREF is used in a remote data mirroring (RDM) system. The model of NFRs and the analysis of their preferences are enabled by using dynamic decision network (DDNs) and Bayesian Surprise.