Flexible and dynamic network coding for adaptive data transmission in DTNs
Existing network coding approaches for Delay-Tolerant Networks (DTNs) do not detect and adapt to congestion in the network. In this paper we describe CafNC (Congestion aware forwarding with Network Coding) that combines adaptive network coding and adaptive forwarding in DTNs. In CafNC each node learns the status of its neighbours, and their ego-networks in order to detect coding opportunities, and codes as long as the recipients can decode. Our flexible design allows CafNC to efficiently support multiple unicast flows, with dynamic traffic demands and dynamic senders and receivers. We evaluate CafNC with two real connectivity traces and a realistic P2P application, introducing congestion by increasing the number of unicast flows in the network. Our results show that CafNC improves the success ratio, delay and packet loss, as the number of flows grows in comparison to no coding and hub-based static coding, while at the same time achieving efficient utilisation of network resources. We also show that static coding misses a number of coding opportunities and increases packet loss rates at times of increased congestion.