A Novel Handover Prediction Scheme in Content Centric Networking Using Nonlinear Autoregressive Exogenous Model

Content Centric Networking (CCN) has been recently proposed as a potential candidate for the future Internet architecture. It allows users to focus on the data they are interested in, rather than having to connect to a specific physical host where the data locates. CCN is claimed to inherently support mobility. While this is true for content receivers, source mobility as well as handovers still remain a challenging tasks in CCN. This paper proposes a novel handover prediction mechanism based on received signal strength (RSS), delay and handover cost using the nonlinear autoregressive exogenous model (NARX). We first introduce a mathematical model using Integer Linear Programming (ILP) to solve the handover decision problem assuming that the mobility scenario is known. This ILP model minimizes the overall cost over the considered time period including the cost of handover and the cost of using a certain network. The ILP model guarantees an optimal handover solution, which is then used to train the NARX. After the learning phase, the NARX can automatically make the handover decision based on RSS and delay information. The performance evaluation shows that our new mechanism can avoid the ``ping-pong" effect usually seen in threshold-based approaches, and also outperforms the threshold-based handover in terms of delay, which results in a better Quality of Service.