Mobile Ad Hoc Networks (MANETs) are temporal and infrastructure-independent wireless networks that consist of mobile nodes. For instance, MANETs can be used as an emergent network for communication among people when a disaster occurred and existing infrastructures failed. In another case, participants of a meeting, conference, or event can also build a temporal informationsharing network over a MANET to exchange their own information each other. Since there is no central server in the network, each node has to find out its desired information (objects) by itself. In a MANET, a source node can communicate with its destination node through a multi-hop path. The path between them is determined by a routing protocol. However, these protocols do not provide the source node with the location of its desired objects. Constructing a mobile Peer-to-Peer (P2P) network over the MANET can support the object search. Some researchers have proposed construction schemes of mobile P2P networks, such as Ekta and MADPastry, by modifying distributed hash table (DHT) schemes which are mainly used in the current Internet for the purpose of file-sharing, video/audio calling, etc. Ekta and MADPastry integrated a DHT-based application-layer routing and network-layer routing to increase search efficiency. Furthermore, MADPastry have proposed a clustering method which groups the overlay nodes according to their physical distance. However, it has also been pointed out that the search efficiency deteriorates in highly dynamic environments where nodes quickly move around. In this thesis, we focus on route disappearances in the network layer which cause the deterioration of search efficiency. We describe the detail of this problem and evaluate quantitatively it through simulation experiments. Since it is difficult to tackle this problem using only network-layer approaches, applicationlayer approaches must be required. We extend MADPastry by introducing a method sharing pointers among nodes in a cluster. Through simulation experiments, we show that the proposed method improves the success ratio of object search up to 40 % compared with MADPastry. In addition, we show that the proposed method achieve higher success ratios of object search compared with MADPastry in various environments such as the where queries are sent at a highly frequent rate or the object popularities follow a Zipf distribution.Finally, we considered the parameter tuning of the routing protocol in network layer.

Kei Takeshita, Improving Success Ratio of Object Search in Highly-Dynamic Mobile P2P Networks, Master's Thesis, March 2008.