P4 programmability for fault-tolerant Software Defined Flying ad-hoc Network

This paper introduces the issue of highly available Software Defined Flying ad-hoc Network for critical real-time applications. FANETs are a particular implementation of VANET (Vehicle Ad-hoc NETworks) where the vehicles, in the specific case, are Unmanned Aerial Vehicles (UAV), or drones. VANETs in general suffer from the mobility of the nodes, which can lead to continuous disconnections and topology changes. This issue becomes particularly troublesome for Flying ad-hoc networks (FANET), which are characterized by flying nodes, with low node density and a high rate of topology changes. On the other hand, nowadays, drones are widely used both in civil and industrial environments, and thus they may need to satisfy certain requirements in terms of the availability of communication. In the traditional distributed networks, the propagation of the information when there is a change in the topology, and the consequential update of the information in each switch, can introduce long periods of unavailability, which cannot be tolerated in certain contexts. For this reason, we decided to introduce an approach based on SND to manage FANET in an efficient and reliable way. The SDN architecture relies on a centralized controller to configure the network devices by exploiting a control plane channel. However, having a centralized controller implies a single point of failure. Moreover, control plane links are critical in SDN, since the controller uses those links to update the configuration of the switches. In order to efficiently introduce the SDN paradigm in a dependable FANET it is worth considering a novel approach, where the controller is responsible for the main routing decisions, but the network is able to operate autonomously and tolerate topology changes, within a certain degree, even when the SDN controller is not connected. The proposed architecture is a P4-SDN, that is capable of automatically rerouting traffic even when the controller can not communicate with the involved devices, providing then an additional level of redundancy. The programmable switches are instructed to take countermeasures when an unexpected link fault occurs and the controller is unreachable. In our solution, the controller configures both primary and backup actions for each expected flow, leveraging the programmability of P4-enabled devices. Thus, each switch can choose to forward the packets over the primary path or follow the backup action. Since these countermeasures are taken locally and independently from the controller, this approach allows the network to operate even when the control-plane link is unusable. It also improves the handover of the fault when the controller is reachable, minimizing critical application traffic disruptions.