Validating Code-Level Behavior of Dynamic Adaptive Systems in the Face of Uncertainty

A dynamically adaptive system DAS self-reconfigures at run time in order to handle adverse combinations of system and environmental conditions. Techniques are needed to make DASs more resilient to system and environmental uncertainty. Furthermore, automated support to validate that a DAS provides acceptable behavior even through reconfigurations are essential to address assurance concerns. This paper introduces Fenrir, an evolutionary computation-based approach to address these challenges. By explicitly searching for diverse and interesting operational contexts and examining the resulting execution traces generated by a DAS as it reconfigures in response to adverse conditions, Fenrir can discover undesirable behaviors triggered by unexpected environmental conditions at design time, which can be used to revise the system appropriately. We illustrate Fenrir by applying it to a dynamically adaptive remote data mirroring network that must efficiently diffuse data even in the face of adverse network conditions.