Bandit Algorithms for Robust Forwarding in Named Data Networking

To meet the current Internet usage, Named Data Networking (NDN) is a promising paradigm that proposes to shift the Internet architecture from host-centric to data-centric. One of the key features of NDN is forwarding, which has a critical impact on network performance. In this paper, we focus on robust forwarding of interest packets under changing network conditions. Based on bandit algorithms, we devise forwarding approaches where the aim is to efficiently balance the exploration of new paths and the exploitation of previously learned paths without relying on the Forwarding Information Base (FIB) table. Simulation results show that our proposal brings significant benefits, such as responding quickly to network congestion and link failures, while reducing the forwarding cost in terms of prefix name lookup and memory requirement in FIB.