Packet size optimization in delay tolerant networks

Studies have shown that the size of the transmitted packet has a significant impact on the performance of wireless networks, especially those in challenged environments such as Delay Tolerant Networks (DTNs) where bandwidth is a precious resource. In order to improve the end-to-end goodput in DTNs, segment size at the convergence layer and bundle size at the bundle layer are jointly optimized in this paper. Different from existing work that only performs experimentations to find the optimal packet size in single-hop DTNs, our work focuses on the theoretic analysis and formulates the relationship between packet size at two layers (i.e., the convergence layer and the bundle layer) and the performance metrics (i.e., delay, goodput) of single-hop and multi-hop DTNs. The analysis not only confirms the existence of the optimal packet size, but also quantitatively shows that the optimal packet size is determined by the number of hops, wireless channel conditions in DTNs. Furthermore, a novel Packet Size Optimization Algorithm (PASO) is proposed, in which segment size and bundle size are jointly adjusted to maximize the goodput in multi-hop DTNs. Compared with schemes where joint packet size optimization is not considered, PASO can achieve average 15% more goodput in typical DTN environments where channel conditions are usually assumed to be poor.