Unveiling the Features of Network Dynamics: a Data-Oriented Approach

The ability to predict network dynamics is a challenging problem for communication providers as this can have an impact on the performance of the applications running on top of their infrastructure. In this paper we use a data-driven approach to investigate whether observing the variation of the connectivity properties of a network over time can provide insights onto the occurrence of network topological changes. In that direction, we propose a methodology to model various patterns of network dynamics and develop a user-configurable software tool enabling the collection of network connectivity metric data for any type of communication network topology and based on a wide range of dynamics patterns. We use our tool to constitute a dataset of network connectivity metric samples based on 126 real network topologies and 15 different types of dynamics patterns. The analysis performed based on the obtained dataset shows that identifying how connectivity metrics vary has potential for characterizing network topological changes.