Autonomic link management in wireless backhaul networks with OpenFlow and traffic prediction

Microwave radio is the predominant technology for backhaul or backbone networks, and its usage is expected to increase further. A way to increase capacity in this type of network is to aggregate multiple physical links into a single logical link. However, power is being wasted when the aggregated capacity is not being fully utilised. There are cost savings to be gained by only having the required number of physical links on. This paper presents a system that utilises extensions of the OpenFlow protocol to dynamically switch on/off physical links to meet capacity requirements while minimising power consumption while the Autoregressive Integrated Moving Average (ARIMA) model is used to predict the traffic load to be carried on aggregated (microwave) links. The proposed system is implemented on commercial-off-the-shelf microwave routers and experimentally validated on a testbed.