Emergent Behavior in Complex Systems Engineering

Chapter 11TOWARD THE AUTOMATED DETECTION OF EMERGENT BEHAVIOR

Claudia Szabo and Lachlan Birdsey

School of Computer Science, The University of Adelaide, Adelaide, SA, 5005, Australia

INTRODUCTION

Complex systems often exhibit behavior that cannot be reduced to the behavior of their individual components alone and require thorough analysis once unexpected properties are observed (Davis, 2005; Johnson, 2006; Mogul, 2006). These emergent properties are becoming crucial as systems grow both in size (with respect to the number of components and their behavior and states) and in coupling and geographic distribution (Johnson, 2006; Mogul, 2006; Bedau, 1997; Holland, 1999). Due to significant research interest in the past decades, a plethora of emergent properties examples, from flocks of birds, ant colonies, to the appearance of life and of traffic jams, has been observed and identified. In software systems, connection patterns have been observed in data extracted from social networks (Chi, 2009) and trends often emerge in big data analytics (Fayyad and Uthurusamy, 2002). More malign examples of emergent behavior include power supply variation in smart grids due to provider competition (Chan et al., 2010), the Ethernet capture effect in computer networks (Ramakrishnan and Yang, 1994), and load-balancer failures in a multi-tiered distributed system (Mogul, 2006). As emergent properties may have undesired and unpredictable consequences (Mogul, 2006; Ramakrishnan and Yang, 1994; Floyd ...