Disaster-resilient virtual-network mapping and adaptation in optical networks

Today's Internet applications include grid- and cloud-computing services which can be implemented by mapping virtual networks (VNs) over physical infrastructure such as an optical network. VN mapping is a resource-allocation problem where fractions of the resources (e.g., bandwidth and processing) in the physical infrastructure (e.g., optical network and servers/data-centers) are provisioned for specific applications. Researchers have been studying the survivable VN mapping (SVNM) problem against physical-infrastructure failures (typically by deterministic failure models), because this type of failure may disconnect one or more VNs, and/or reduce their capacities. However, disasters can cause multiple link/node failures which may disconnect many VNs and dramatically increase the post-disaster vulnerability to correlated cascading failures. Hence, we investigate the disaster-resilient and post-disaster- survivable VN mapping problem using a probabilistic model to reduce the expected VN disconnections and capacity loss, while providing an adaptation to minimize VN disconnections by any postdisaster single-physical-link failure. We model the problem as an integer linear program (ILP). Numerical examples show that our approach reduces VN disconnections and the expected capacity loss after a disaster.