Collaborative strategy learning for distributed network self-configuring

Self-configuration is one of the most important functions of autonomic networks because it determines optimal use of resources during network's operation. However, this task is very complex as it must be performed according to service contracts between users and operators, network's infrastructure and workload. Knowledge Plane is a recently proposed concept to address this complexity by using cognitive tools (learning and reasoning). In this paper, we propose a Knowledge Plane including a distributed and collaborative machine learning method based on inductive logic programming (ILP). The main objective is to achieve distributed self-configuring by learning collaboratively best configuration strategies. We apply it in a practical context (DiffServ) and evaluate effects of this proposal on network's performances and occupation rate.