LBHD: Drop policy for routing protocols with delivery probability estimation

Abstract

Delay Tolerant Networks (DTN) are characterized by a lack of end-to-end connectivity. As such, messages (called bundles) can be stored in buffers for a long time. Network congestion can result in poor delivery rates, as bundles are dropped before having a chance of reaching their destination. Some routing protocols, such as the Probabilistic Routing Protocol using History of Encounters and Transitivity (PRoPHET), maintain estimations of delivery probabilities for each destination. In this paper, a new drop policy called Largest Bundle's Hosts Deliverability (LBHD) is proposed that considers all the hosts that received a replica of the same bundle, and their respective delivery probability as estimated by PRoPHET. LBHD uses this additional information to better manage congestion. Simulation results show that LBHD achieves the best delivery probability when compared with other drop policies proposed in the literature. The LBHD policy achieved an average delivery rate 37% better than the default Drop Head (FIFO) policy at the price of a 21% increased average delay.