Effective Decision Making in Self-adaptive Systems Using Cost-Benefit Analysis at Runtime and Online Learning of Adaptation Spaces

Abstract

Self-adaptation is an established approach to deal with uncertainties that are difficult to predict before a system is deployed. A self-adaptative system employs a feedback loop that tracks changes and adapts the system accordingly to ensure its quality goals. However, making effective adaptation decisions at runtime is challenging. In this chapter we tackle two problems of effective decision making in self-adaptive systems. First, current research typically focusses on the benefits adaptaton can bring but ignores the cost of adaptation, which may invalidate the expected benefits. To tackle this problem, we introduce CB@R (Cost-Benefit analysis @ Runtime), a novel model-based approach for runtime decision-making in self-adaptive systems that handles both the benefits and costs of adaptation as first-class citizens in decision making. Second, we look into the adaptation space of self-adaptive systems, i.e. the set of adaption options to select from. For systems with a large number of adaptation options, analyzing the entire adaptation space is often not feasible given the time and resources constraints at hand. To tackle this problem, we present a machine learning approach that integrates learning with the feedback loop to select a subset of the adaption options that are valid in the current situation. We evaluate CB@R and the learning approach for a real world deployed Internet of Things (IoT) application.