Ε-time early warning data backup in disaster-aware optical inter-connected data center networks

Backup in data center networks (DCNs) against disasters is a critical task for avoiding huge data loss. In this paper, we optimize data backup for a particular DCN node threatened by a disaster by assuming geodistributed optical interconnected DCNs, where the node can be aware of the disaster in an E time before it is disrupted. We first formulate an integer linear program (ILP) to find the maximum amount of data in the threatened DCN node that can be protected. This helps to determine which data should be protected according to data importance. Then we formulate another ILP to achieve minimum-cost backup by properly selecting a set of safe backup DCN nodes and corresponding backup routes. To get real-time solutions for engineering practice, we also propose a heuristic to achieve cost-efficient backup in ε-time early-warning disasters. Extensive numerical results show that the proposed algorithms can automatically adapt to different early warning times E for generating cost-efficient data backup solutions.