Assessment of Configuration Stability and Variability in Collections of Self-Adaptive Systems

Abstract

Self-adaptive and self-organising (SASO) systems are typically composed of several (semi-)autonomous subsystems that alter their configuration and the system's structure in response to environmental and internal observations. The overall goal of these changes is mainly two-fold: improving the expected performance or goal achievement and providing increased robustness against disturbances and unforeseen events. In order to establish self-awareness of overall system behaviour as a basis for guided control intervention, we investigate measures to quantify and assess system properties. This paper introduces a novel approach to determine a degree of configuration variability and stability based on external, system-wide observation of configuration variables of the distributed subsystems. We analyse the behaviour of our metrics in two different scenarios and outline the possible advantages of the applicability.