Improving Priority-Awareness Of Non-Functional Requirements During Decision-Making In Self-Adaptive Systems

Abstract

Self-adaptive systems (SASs) exhibit autonomous decision-making to deal with uncertainty in their operating environments. A fundamental problem with SASs is to ensure that their requirements remain satisfied as they adapt. Trade-off analysis of the non-functional re-quirements (NFRs), based on their satisfaction priorities, is a key to establishing a balance among them. Such trade-off analysis is often based on optimization techniques comprising decision analysis and utility theory. A problem with these techniques is that they use a single-scalar utility value to specify a combined priority for all the NFRs. Nevertheless, this combined priority does not give any information about the impacts of the environmental contexts on the individual priorities of NFRs. Moreover, these separate NFR priorities may change according to the runtime environmental contexts. Therefore, there is a need to have an approach that supports the runtime, autonomous reasoning with the distinct priorities of NFRs during the decision-making process. This PhD thesis addresses this problem by presenting Pri-AwaRE, a self-adaptive architecture for decision-making in SASs. The approach uses Multi-Reward Partially Observable Markov Decision Process (MR-POMDP) as a runtime specification model to support the modelling of the individual NFRs’ priorities. The MR-POMDP model also provides runtime reasoning and autonomous tuning of these separate priorities. Therefore, it underpins priority-aware decisions. The approach has been evaluated using two substantial case studies from the different networking domains. A comparison with other state-of-the-art approaches has also been carried out. The results have shown that the priority-aware decisions offered by Pri-AwaRE provide compliance with the requirements for both the case studies even under the changing environmental contexts at runtime