Analysis of multiple-unicast throughput in finite-buffer Delay-Tolerant Networks

The problem of computing the throughput capacity of general unicast in a Delay-Tolerant Network under the finite-buffer regime is addressed in this paper. A sparse mobile wireless network deployed on an M times M square-grid is considered, wherein m unique mobile source/destination pairs communicate with the aid of n mobile relay nodes using the store, carry, and forward paradigm. Each mobile relay node is equipped with a finite storage. Thus, several source/destination pairs contend for limited network resources. Under this setup, the throughput achieved per source-destination pair at steady-state node-mobility is analyzed, incorporating practical considerations such as node-to-node contention and finite communication range, for a class of two-hop relay protocols. In addition, two different approaches for buffer management are considered. The paper shows that, using a novel approach based on embedded Markov-chains, accurate analysis of the throughput can be achieved for the above model, as it is validated by simulations. A significant conclusion of this work is that considerable throughput improvements can be achieved by judicially managing the relay-node buffer-space.