Abstract
This paper investigates optimum distributed storage techniques for data preservation, and eventual dissemination, in opportunistic heterogeneous wireless sensor networks where data collection is intermittent and exhibits spatio-temporal randomness. The proposed techniques involve optimally sharing the sensor nodes' storage and properly handling the storage traffic such that the buffering capacity of the network approaches its total storage capacity with minimum energy. The paper develops an integer linear programming (ILP) model, analyses the emergence of storage traffic in the network, provides performance bounds, assesses performance sensitivities and develops quasi-optimal decentralized heuristics that can reasonably handle the problem in a practical implementation. These include the Closest Availability (CA) and Storage Gradient (SG) heuristics whose performance is shown to be within only 10% and 6% of the dynamic optimum allocation, respectively. Copyright (c) 2016 John Wiley & Sons, Ltd.