Optimal network structuring for large-scale WSN with virtual broker based publish/subscribe

Abstract

Large-scale wireless sensor networks (WSNs) often need to be partitioned for efficient data acquisition. To enhance the scalability and reliability of publish/subscribe (pub/sub) systems implemented on a large-scale in WSNs, dedicated broker devices with large storage and computation capacities can be replaced with virtual brokers (VBs), each formed by multiple co-located sensor nodes that shared resources to collaboratively perform the role of a conventional broker. As communication is a major consumer of energy, which is a scarce resource in WSNs, how the WSN is structured in terms of the number of partitions and the size of VB can impact the energy consumption of the network. In this paper, we propose a novel approach to optimal network structuring of large-scale WSN with virtual brokers for pub/sub communication. Underlying the approach is an analytical model which determines the optimal number of partitions and VB size that jointly minimize the total communication overhead of the WSN.