DRL-based Forwarding Strategy in Named Data Networking

Abstract

Named Data Networking(NDN) is one of Information-Centric Networking that have a complete routing protocol framework and hierarchical naming rules. In NDN, the name of each interest packet is unique, and the name of the interest packet contains much information of requested content. The NDN router forwards the interest packet according to the forwarding strategy after finding the Forwarding Information Base(FIB). In this paper, we combine Deep Reinforcement Learning(DRL) and forwarding of NDN to propose an intelligent forwarding strategy. We gather information, such as the data content, face states and network states of the interest packet in the forwarding process and save these features as inputs of the DRL for training, whose result will be used as the forwarding strategy to guide the forwarding of interest packet. Experimental results show that the proposed forwarding strategy effectively reduces Round Trip Time(RTT) by 7.7% and improve the throughput compared with the existing forwarding strategy.