Anycast (re)routing of multi-period traffic in dimensioning resilient backbone networks for multi-site data centers

Abstract

We consider the problem of dimensioning resilient backbone networks for cloud-like scenarios where demand is to be served at one among several candidate data centers (DCs), and where that demand varies over time, which we assume to be slotted. We thus consider multi-period traffic, for which we need to find routes to both a primary DC and a backup DC (in case the primary, or the network connection towards it, fails) and account also for synchronization traffic (following its own routes) between the two chosen DCs. We propose a path formulation and adopt a column generation approach: the (restricted) master problem (RMP) selects “configurations” to use for each demand in each of the time periods, while pricing problems (PPs) construct new, potentially cost-reducing configurations for a given demand. Our model allows for several PPs to be solved in parallel, and we demonstrate the time savings achieved by doing so. We compare several anycast (re)routing strategies, where we allow traffic that spans multiple periods to either (i) not be rerouted in different periods, (ii) only change the backup DC and routes, or (iii) freely change both primary and backup DC choices and routes towards them.