Message Dissemination in Delay-Tolerant Networks with Probabilistic Encounters

Abstract

For some applications of Delay-Tolerant Networks (DTNs), such as publish/subscribe systems, messages do not have explicit destination addresses. In those cases, message routing is driven by message content instead of by destination addresses. In this paper, to achieve more efficient and robust message dissemination, a Content Encounter Probability based Message Dissemination (CEPMD) scheme for DTNs is proposed. In CEPMD, we use request predicates to represent subscribers' interests and propagate request predicate messages over the network. A request predicate is tagged with an ep, representing the probability that the predicate's host node will encounter the destination node. Content messages are forwarded to the nodes that have a request predicate that matches the content message and have a greater ep value. The corresponding content messages can be forwarded to the subscriber along the probability gradient of predicates. The experimental results show that CEPMD outperforms the state-of-the-art DTN algorithms on both message delivery ratio and delivery delay.