On the performance and policies of mobile peer-to-peer network protocols

Certain "data diversity" approaches have been proposed recently as the means to avoid the demonstrated lack of scalability in mobile ad-hoc networks (MANETs). However, exploiting data diversity requires the development of particular protocols that incite the cooperation of nodes. A feature of the protocols we propose is that, instead of assuming all nodes to be interested in the same data item, we assume different per-node data items of interest and a large population of data items (files) with a popularity that follows a Zipf distribution. We also consider dynamic environments where, over time, new nodes join the network (and others depart). Together with a one-for-one reciprocal exchange policy, we observe the impact of density and per-node initial file sets. Extending the basic idea of adjacent node exchanges to a limited-length path routing, and taking care of proper incentives for nodes to route data on behalf of other nodes, we find an improved throughput for the entire system. However, we find no clear evidence that the per-node perceived throughput for a node's desired files is any different than in the non-cooperative case. Thus, we question whether there exist incentive schemes that could engage nodes to support the data diversity paradigm.