Abstract
The evolving spreading of data-centres' implementations with hundreds of thousands of servers in-site, has imposed new benchmarks on network survivability. Therefore, assessing the network availability under a set of geographically correlated failures known as regional failures has become imperative. In this paper, we assess the network robustness against regional failures. In particular, we develop a probabilistic model to estimate the availability of a network's link in the event of a failure scenario. Then, using the estimated probabilities of all links to provide a general overview of the network-wide availability. This paper is distinguished by providing a failure model that is continuous over the failure region. This model provides a quantitative tool to estimate the amount of resources required for repairing the faulty parts of the network. Finally, extensive numerical results on the US backbone network for different failure scenarios indicate the applicability and accuracy of the proposed model. Byline: Mohamed Mostafa A. Azim