MDP-based forwarding in Named Data Networking

In Named Data Networking (NDN), a content can be replicated in multiple network node. Thus, more than one interface can be selected for the same content in a node, and it is very important to choose the interface to forward request packet for better network performance. In this paper, we introduce a MDP-based forwarding strategy which is under the consideration of network problem and location of content. The whole scheme consists of three segments. First, we develop a Markov Decision Process (MDP) model for the whole forwarding process in a single node, and utilize MDP to solve the forwarding decision problem. Second, we build the request output queue based on content and utilize queuing theory to estimate the real-time network status, which is used as immediate reward in MDP. Last, a rate control mechanism is provided to solve the rate issues caused by the fast forwarding of request packet, and it is used to built the stable queue. Numerical experiments within a set of network set are implemented by ndnSIM, and the result of experiments demonstrates a good performance of low delay.