Adaptive Interest Modeling Improves Content Services at the Network Edge

Abstract

Content-based mobile ad-hoc networking (MANET) can be rapidly deployed in emergency and tactical situations at the network edge where fixed infrastructure is lacking. How to get the relevant content to the right user quickly, in the midst of network disruptions and resource constraints, is a key research challenge. Our insight is that if the network has some notion of users’ interests and information needs, it can take proactive actions in order to make the network serve the users better. We have developed adaptive interest modeling (AIM) to capture and model user information needs at the network layer. AIM works by continuously monitoring network events at each node, and incrementally processing them to build an interest model (IM) for the node. One key contribution of our approach is to anchor AIM at the content-based network layer, which allows all upper level mobile applications to benefit without modification. This adaptive IM can enable many user-aware features including content prefetching and IM sharing. Our unique IM enabled prefetching is based on recognizing user situations. IM sharing is a novel and efficient way of automatically keeping users up to date with each other in tactical scenarios. We implemented AIM as well as the IM-enabled prefetching and IM sharing features in a content-based MANET prototype. We evaluated these features in a network emulation environment. IM-enabled prefetching significantly reduces response time while increasing data availability at the same time. When combined with IM sharing, additional sizable reduction in response time is achieved.