Network failures support for traffic monitoring mechanisms in software-defined networks

© 2016 IEEE. Nowadays, monitoring per-flow metrics, such as throughput and delay, in OpenFlow networks represents an important input to implement traffic engineering approaches. Per-flow monitoring mechanisms require to uniquely identify the complete network path followed by all the packets of the corresponding flow. For example, in order to measure path delay, monitoring mechanisms regularly inject probe packets at the first switch of the path, such that these packets travel exactly along the same path of the corresponding flow till the last switch, which is in charge of sending them back to the network controller to compute the path delay. However, while a flow is active, any of the switches of the corresponding path can suffer different kinds of failure. In this situation, a new path must be recomputed for this flow by the controller in a timely manner, and the monitoring tool must be also updated. In this paper we present a mechanism which improves the behavior of per-flow monitoring modules allowing them to react in the face of network failures.