Post-failure repair for cloud-based infrastructure services after disasters

Cloud infrastructure services allow organizations to outsource their computing, storage, and networking needs to external providers. These offerings use network virtualization to provision customized, interconnected resource pools across substrate infrastructures comprised of datacenter computing/storage systems and interconnecting networking switches/links. Now as virtual infrastructure services expand, disaster recovery concerns are coming to the fore. However, only a few studies have looked at network virtualization survivability for large-scale correlated multi-failure disaster scenarios. Since most existing efforts here focus on protection methods to pre-provision backup resources, they cannot ensure recovery from generalized multi-failure events. Hence post-failure recovery is a critical concern, particularly since damaged infrastructures will be repaired in a time-staged, incremental manner. Along these lines, this paper presents a comprehensive framework for progressive recovery schemes for network virtualization. Namely, a range of efficient heuristic and metaheuristic resource placement strategies are defined and analyzed. Overall results show significantly improved recovery rates and speed with distributed load-based schemes.