Building reliable systems based on self-organizing multi-agent systems
Florian Klein, Matthias Tichy · 2006
High reliability and availability are of utmost importance to the majority of today's software-based systems. Typically, some kind of redundancy is used in order to achieve high reliability and availability in case of faults. Dynamic redundancy setups employ backup systems which take over as soon as the primary system fails. Additionally, central fault tolerance management systems are often used to guide the replication process. In this paper, we present an approach which relies on emergent fault-tolerance. The desired behavior is the result of local reconfigurations of self-organizing agents. This allows us to abandon the classical distinction of primary and backup systems as well as the need for a central management system. Our experimental results show the benefit of our approach.