LAUSR

Network failure caused by disasters (both natural and man-made like earthquakes, floods, cyclones, electromagnetic pulse attacks etc.) result in communication disruption and huge amounts of data loss in the backbone datacenter (DC) networks. To prevent such large-scale network disruptions and quickly resume connectivity after the disaster, network operators require improved and efficient data-transfer algorithms in geographically distributed (geo-distributed) optical inter-DC networks. Minimising loss of infrastructure and preventing network disruption requires estimating the damage from a possible disaster. In this study, the authors consider a mutual backup model, where DCs can serve as backup sites of each other, thereby significantly reducing the backup duration (i.e. DC-Backup-Window (DC-B-Wnd)). They specifically consider the joint optimisation of probabilistic backup site selection and the amount of data to be backed up. They propose mixed-integer linear programming models for backup time minimisation using a single DC as well as dual DCs at backup sites. Further, they investigate the trade-off between DC-B-Wnd and the computational complexity of the proposed algorithms and perform extensive numerical simulations to show that, in the case of disasters, single and dual DC backups with risk-aware probabilistic path selection give shorter backup windows as compared to existing algorithms.