• Finite State Machine
• NETCONF
• network virtualization
• Proactive recovery
• qemu
• SDN

Emerging services applications (such as industrial automation and autonomous driving) require efficient control and management of the network infrastructure in order to assure stringent requirements (e.g., in terms of delay and jitter). Software-Defined Networking (SDN) provides efficient network control; however, some issues should be deeply investigated: e.g, scalability, failure recovery and real-time responsiveness. Fast network recovery when a fault occurs is not easy to achieve in large networks, due to the processing of the huge number of restoration requests to the SDN controller, resulting in a large recovery time. However, previous work demonstrated that dynamic fault recovery can be improved using proactive recovery mechanisms. Among them, a recovery mechanism based on a finite state machine (FSM) is here assumed operating with NETCONF protocol. While the network is operating in normal conditions, with FSM-based recovery, the SDN controller periodically instructs the network agents (i.e., devices’ local controllers) about the actions to perform in case of failure, which leads to faster recovery to failures. This way, when a failure occurs the agents react without SDN controller intervention. Such an approach is here investigated for Time-Sensitive Networks (TSNs).
This thesis first provides a literature review of i) proactive and reactive recovery strategies in SDN networks; ii) NETCONF protocol and YANG data modelling; iii) TSN. Then, it presents the FSM-based recovery controlled through NETCONF and YANG, named Delegated Restoration for Time-Sensitive Networks (DRTSN). Details on the implementation of DRTSN are here provided spanning from data to control plane. Finally, measurements are performed to evaluate the performance of DRTSN.