Maximizing Last-Minute Backup in Endangered Time-Varying Inter-Datacenter Networks

Abstract

Natural disasters are time-varying in nature. They adversely affect backbone datacenter (DC) networks, thereby resulting in huge loss of data within a short span of time. Maximizing last-minute data backup in an endangered DC network hit by a progressive disaster is, therefore, of utmost importance to ensure data protection and service continuation. Thus, in this article, we propose a novel Mixed-Integer Linear Program (MILP) and a heuristic (D-RADDAR) with reduced time-complexity to address risk-aware emergency backup in endangered DC networks hit by dynamic disasters. The mathematical framework for maximizing the overall last-minute backup percentage is developed. A quantitative analysis of the risk associated with such last-minute backups is carried out. The proposed algorithms are evaluated in terms of the DC backup success percentage and compared to existing dynamic backup algorithms in the literature. Extensive simulations indicate that among all existing heuristics, the D-RADDAR achieves maximum data evacuation while maintaining minimal risk and having a polynomial time-complexity.