Multi-agent Reflection in Autonomic Systems

Increasing complexity and scales of computing systems bring growing demands on their designers and administrators. These issues have been explored from different viewpoints. While the multi-agent system researchers focus on the coordination and theoretic properties of distributed system, the researchers from the field of autonomic computing focus on the deployment of self-adapting systems in real domains. This contribution tries to bring together both domains and presents the multi-agent reflection from the autonomic computing point of view. We show that proposed architecture of reflective agent covers all the characteristics of an autonomic computing system and furthermore, we describe some properties of reflective agents that can further extend the autonomic computing system, especially its self-optimizing abilities. All the characteristics of autonomic computing system are demonstrated on three use cases, where the reflective agents have been deployed, together with the results of experiments.