Self-adapting network topologies in congested scenarios.

Most studies in complex networks assume that once a link is created between two nodes it is never deleted. However, there is a recent interest towards systems where links can be rapidly rewired. An important issue in that type of networks is to discover the topology that, given a search algorithm, optimizes the search process. In this paper, we present a system model that, depending on the current network congestion, makes nodes to establish link connections so that the resulting topologies tend to a starlike topology when congestion is small and to randomlike topologies when congestion becomes relevant. Those topologies have been shown to be optimal in the above-mentioned conditions. Such a model can be easily implemented in practice and therefore, may be relevant in areas as the topology management of peer-to-peer networks.