Scalable decentralized control for sensor networks via distributed lattices

• Authors: Baruch Awerbuch;Jonathan Stanton
• Venue: IPSN'03 Proceedings of the 2nd international conference on Information processing in sensor networks
• Year: 2003

A network of embedded devices needs to be able to execute queries for dynamically changing content. Even in a completely reliable network, this is a formidable task because of the enormous scale of the networks, severely limited resources of individual devices (bandwidth and battery power) and the heterogeneity of resources being managed. In this work we introduce a novel information query methodology for designing online solutions for heterogeneous sensor networks with various resources (e.g., battery, bandwidth, CPU). This provides a route selection and query management mechanism that will enable a sensor network to find sensor level information without a routing algorithm specialized for the particular form of information. In order to execute such a methodology in a scalable, limited resource environment such as sensor networks we employ a novel lattice data structure, which is basically a combination of trees with small overlap that provably enables extension of any routing or directory infrastructure to an arbitrary scale, with only small overhead. We show how to use such data structures (lattices) that will enable scaling to millions of devices with an overhead that only grows logarithmically in the number of network nodes and with provably small distortion of paths. Moreover, we show a completely distributed implementation of such structures, that creates minimal overload on the client sensors.