Scaling properties of delay tolerant networks with correlated motion patterns

Abstract

Mobile wireless networks with intermittent connectivity, often called Delay/Disruption Tolerant Networks (DTNs), have recently received a lot of attention because of their utility in various application scenarios where delay is noncritical. DTN routing and transport protocols effectively overcome partial connectivity by letting the nodes carry-and-forward data. The scalability of DTN protocols is very important for protocol design and evaluation. In particular, we need models that allow us to predict the performance of DTNs as a function of node mobility behavior (e.g., inter-contact times). Yet so far little work has been done to develop a unified framework that formalizes DTN performance as a function of motion behavior. In this paper, we represent DTNs as a class of wireless mobile networks with intermittent connectivity, where the inter-contact behavior of an arbitrary pair of nodes can be described by a generalized twophase distribution consisting of a power-law head with an exponential tail, which represents correlated node mobility. Recent experiments have confirmed that such a two-phase distribution is a more realistic model for real traces collected from vehicular and pedestrian scenarios than the previous models based on random mobility and Poisson assumptions. Using this DTN model, we make the following contributions. First, we extend the throughput and delay scaling results of Grossglauser and Tse (originally derived for an exponential inter-contact time distribution) to a more general mobility model with a two-phase distribution. Second, we analyze the impact of finite buffer on the capacity scaling properties of DTNs, again for different correlation behaviors. Finally, we validate our analytical results with a simulation study. Copyright 2009 ACM.