Performance of wireless sensor networks under random node failures

Networked systems are essential to the function of modern society and the consequences of damage to networks can be severe. Assessing the performance of a network is an important step for recovering damaged networks and designing reliable networks. Some of the key general indicators of network performance are connectivity, distance between node pairs, and number of alternative routes. We focus on sensor networks with a topology modeled by a class of random geometric graphs (RGGs). In order to evaluate survivability and reliability, we consider two types of failure modes in a RGG: uniform and localized node failures. Since network performance is multi-faceted, and assessment can be time constrained, we introduce four measures, each of which can be computed in polynomial time, to estimate performance of a damaged RGG. Theoretical analysis of these four measures is challenging, especially when the underlying graph becomes disconnected. The focus of this paper is to conduct simulation experiments on several measures of network performance through the temporal process of node failures. Together with the empirical results the performance measures are analyzed and compared in order to provide understanding of the two different failure scenarios in a RGG.