Infective flooding in low-duty-cycle networks, properties and bounds

Abstract

Flooding information is an important function in many networking applications. In some networks, as wireless sensor networks or some ad-hoc networks it is so essential as to dominate the performance of the entire system. Exploiting some recent results based on the distributed computation of the eigenvector centrality of nodes in the network graph and classical dynamic diffusion models on graphs, this paper derives a novel theoretical framework for efficient resource allocation to flood information in mesh networks with low duty-cycling without the need to build a distribution tree or any other distribution overlay. Furthermore, the method requires only local computations based on each node neighborhood. The model provides lower and upper stochastic bounds on the flooding delay averages on all possible sources with high probability. We show that the lower bound is very close to the theoretical optimum. A simulation-based implementation allows the study of specific topologies and graph models as well as scheduling heuristics and packet losses. Simulation experiments show that simple protocols based on our resource allocation strategy can easily achieve results that are very close to the theoretical minimum obtained building optimized overlays on the network.