Self-Evolving Network Software

Future network systems with myriads of elements must become selfmanaged: Beyond autoconfiguration, they will have to evolve by themselves as context and elements change and to manage their code base by themselves in a self-organized way. Building such autonomic communication networks requires a radical shift in current network technology. In this paper we discuss the issues involved in network software self-repair and self-optimization with regard to automated network evolution. We survey current work in this area and point towards new research directions like autonomic communications. We then introduce the concept of autocatalytic software for computer networks that is based on a chemically-inspired execution model. In this framework, software is constantly regenerated as it runs, permitting to evolve by itself for reasons of self-repair, self-optimization as well as for innovation.