Trustworthy SDN Control Plane for Prioritized Path Recovery

Sammanfattning: Software Defined Networking (SDN) has gained popularity and attractiveness in the past years’ thanks to its dynamic and programmable nature. The possibility to decouple the data plane and control plane allows for the implementation of Internet networks in an innovative way. Thanks to its ease in changing flow rules in network switches, SDN allows network resources optimization. In the case of critical applications, an essential aspect is to ensure connectivity on the network even in case of link failures. Even when a failure causes an interruption of connectivity, the challenge also stays in recovering as fast as possible. Nonetheless, the SDN controller should have the policy to decide which pairs of end-hosts to disable connectivity when there is a shortage of resources to keep the most important connections active. In this thesis, we developed a proactive-reactive SDN controller coded in Python that copes with restoring end-hosts connectivity as fast as possible. The controller prioritizes the couples of end-hosts that need connectivity based on their importance. During a shortage of network resources, the connectivity of pairs of end-hosts with low importance is disabled, and the connectivity between the most important couples can be ensured. We tested our solution with a reactive-only SDN controller and a proactive-reactive SDN controller that does not consider any prioritization order between end-hosts connectivity. Both the benchmark SDN controllers were developed in the thesis. Experiments were run on the same network topology, with the same couple of endhosts involved. The comparison between the proactive-reactive and reactive-only controllers showed the first one to be faster in restoring the connectivity after a failure. It saves time restoring the connectivity and has fewer packets lost under certain conditions in the relationship between the switch-to-switch and the switchto-controller transmission delay. The comparison between the proactive-reactive iii controller and the controller with no prioritization confirms that without an ordered queue of priorities, it may be the most important couple of end-hosts to lose connectivity in case of shortages of network resources. To simulate a realistic scenario, the project considers the case study of electric power transmission networks using SDN. In particular, the focus is on reconnecting Phasor Measurement Unit (PMU)s to the power grid to ensure system observability. During our experiments, we adopted the typical measurement transmission frequency used by PMUs (50Hz). The SDN switches are deployed with P4, and the SDN controller is coded in Python. Furthermore, it exploits P4Runtime to communicate with the switches in run-time.