Locality Sensitive Information Brokerage in Distributed Sensor Networks

In sensor network applications, sensors often need to retrieve data from each other. Information brokerage is a scheme that stores data (or index files of data) at rendezvous nodes, so that every sensor can efficiently finds the data it needs. A very useful property for information brokerage is locality sensitivity, which means that a sensor close the original source of the data should also be able to retrieve the data with a small communication cost. Given the locality sensitivity requirement, the key is to design an information brokerage scheme that minimizes the storage cost. In this paper, we present a locality sensitive information brokerage scheme. It is designed for general locality-sensitive requirements, which include the linear data-retrieval cost (a frequently studied scenario) as a special case. We also prove that for a large class of networks, in the scenario of linear data-retrieval cost, our scheme achieves the asymptotically optimal storage cost. The result also proves the optimality of a few other schemes in the literature.