Realizing self-management via self-optimization in dynamic networks

Current evolutions of telecommunication networks largely highlight the need for a full exploitation of network capabilities. In parallel, operators face the challenge of preventing the explosion of their operational costs, while already struggling to adapt network configuration to next-generation services. As a result of the ongoing changes in the service/traffic profile, future networks will be characterized by an increasing degree of dynamicity and complexity, as long as by an additional scaling in size, so requiring regular adjustments of network configuration. This task must be accomplished in a flexible, automated and accurate way, enough to abandon the over-provisioning paradigm that is today the preferred solution to face dynamicity in network behavior. It is a shared belief that self-management is the right framework to address all these network operation issues. More precisely, the focus of the present paper is on self-optimization, which is one of the four facets of self-* systems ([14]). Within networking context, optimization problems share the objective of efficiency versus both dynamicity and scalability. Self-optimization algorithms appear as an innovative solution for addressing and solving such concerns. Furthermore, self-optimization pleads for embedding