From Reactive to Predictive Flow Instantiation: An Artificial Neural Network Approach to the SD-IoT

The Software Defined Networking paradigm (SDN) is recognized as one of the main forces that can simplify the management of wired and wireless networks. Most of the effort, so far, has been on the development of the tools and methodologies that allow network administrators and researchers to implement the desired behaviors in SDN networks. A further step in the simplification of the management is the design of a framework that can automatize the control of the network by leveraging the information gathered at the SDN controller combined with machine learning techniques. In this paper a solution is introduced that learns and predicts recurrent patterns in network load and decides which are the best policies to be implemented in order to reach a given objective. To test the proposed solution, a tool has been implemented that adapts the behavior of a wireless sensor network to reach a trade-off between energy consumption and fairness in a simulated environment that exploits the dataset produced by the SmartSantander testbed.