Dynamic probabilistic network protection in large-scale failure scenarios

Abstract

Large-scale failure resulting from natural disasters or intentional attacks is now considered a serious risk for communication network infrastructure. In these events, simultaneous damages in several links and nodes may cause substantial loss of information, which can be very costly for governments, subscribers and businesses. The impact of natural disasters generally is probabilistic in nature. Geographical characteristics and the distance of the components to the centre of the disaster may change the failure probability. Considering the probabilistic failure feature in natural disasters and the possible vast area coverage, we aim to develop a probabilistic dynamic model to protect data from failure and maintain undisrupted network services in large-scale failure scenarios. For this purpose, we develop a preventive protection model, which is able to estimate the potential destruction of all the network components in different locations. Using this information, the proposed model has a holistic view of the failure probabilities for the different paths to make a decision to reroute traffic from the endangered routes through the more reliable paths prior to the failure. As the proposed model protects data before failure, the size of damaged traffic will decrease and fewer connections need to be restored. The proposed preventive model is able to adjust rerouting decision parameters in a dynamic way by considering the disaster expansion and available network resources at each decision interval. Our findings show that the proposed preventive protection model significantly reduces the average number of disrupted connections and successfully decreases the required network restoration time. The performance of the proposed model has been examined in software defined networking (SDN), which is one of the emerging technologies in communication networks. We studied the performance of a SDN controller instructed with a considerable amount of data flow updates and the best method of applying preventive rerouting is indicated.