Optimal two-hop routing in delay-tolerant networks

In this work we deal with the problem of dynamic decentralized self-optimization of forwarding protocols in delay-tolerant networks. We consider the two-hop forwarding protocol and we model the routing problem as an optimization one. We describe, under mild assumptions on the contact process among mobile nodes, the connection between the number of copies of a message and the cumulative probability distribution function of the message delay. We show that, despite the possible presence of a plurality of optimal policies, there always exists a threshold policy that is optimal. We also show that such an optimal strategy can be achieved at run-time by probing the network and with minimal message overhead. The approach is validated by implementing a self-optimizing algorithm in a network simulator; the performance of our solution is analyzed in the case of various real-world contact traces.