On Centrality-Related Disaster Vulnerability of Network Regions

Abstract

Networks are typically embedded in non-homogeneous areas and different parts/regions of the network may therefore be at risk from different types of disasters. This non-homogeneity leads to difficulties in protecting the network against (the risk of) disasters. Network operators need to be able to integrate predictions on possible future disaster events in the planning of their network operation. Especially the (future) availability of network links is crucial in configuring network connections, since the requested availability of network connections is stipulated in Service Level Agreements and must be satisfied, even under the threat of disasters. In this paper, we propose (1) a novel model to characterize disaster areas, with occurrences of each type of disaster represented by a temporal distribution (e.g., Poisson process), and (2) two metrics, namely a betweenness-centrality metric for network regions and an impact metric that indicates the magnitude of the threat posed by disasters within a network region during a given time period.