Brownian gossip: exploiting node mobility to diffuse information in ad hoc networks

Abstract

Several network services, including routing, resource-discovery, etc., require the knowledge of a node's location, before invoking their respective algorithms. Locating a node is a challenging problem in large-scale ad hoc networks, especially when the nodes are mobile. Initial approaches to learn a node's location relied on flooding a query in the network. In view of the large overheads of flooding, gossiping was later proposed as a scalable alternative. With gossiping, each node probabilistically forwards the query. Research findings show that for carefully chosen gossip probabilities, the query is very likely to reach the destination. While gossiping is a random propagation of a specific information, a similar effect might be achieved if nodes move randomly carrying the same information in their local caches. Thus, a node's location information can diffuse into the network via node mobility, as opposed to wireless transmissions that incur bandwidth. If these mobile nodes lend their caches to be queried by its neighborhood, a distributed location estimation service can be envisioned. Clearly, wireless transmission and physical mobility appear to be two modes of information transportation. While each have been studied individually, there has been little work on the possibility of combining them. For example, mobile nodes that gossip with each other periodically, may achieve percolation at significantly lower overheads. This paper investigates the combined potential of gossip and random (Brownian) node mobility. We argue that using such a combined strategy, it might be possible to design a scalable location service for mobile networks, over which several other applications can be developed, including geographic routing, resource discovery, etc