On the role of learning in autonomic network management: The UniverSelf project approach

Acknowledging the shortcomings of traditional network management paradigms, many initiatives attempting to realize and foster the adoption of autonomic network management have taken place while others are still ongoing in the research community. The role of learning is undoubtedly of prime importance in autonomic network management since in principle, it allows for the introduced management and control functionality to exhibit “human”-like self-x attributes and behaviour and adapt to changing network and user conditions. This means that the human operator is offloaded from configuration and reconfiguration tasks that would otherwise require their constant attention and their very deep understanding of low-level network operations and dynamics. When so, human intervention is needed only when, despite the self-x efforts of the various management and control processes, the overall targets cannot be met. In this paper we present the approach of the ongoing UniverSelf project with respect to the role of learning. This paper presents how the learning framework fits in the overall management framework put forward by the project and what the requirements that the learning solutions must exhibit are. Finally, a specific case study corresponding to an instantiation of the learning framework to better illustrate these notions is also described.