A multi-step model for migration and resource reallocation in virtualized network infrastructures

Abstract

In this paper we introduce a novel virtual network migration optimization model which involves multiple steps for the migration process. The model is based on a virtual network embedding (VNE) problem, i.e. the mapping of virtual networks onto a capacitated substrate network [3]. Due to parameter changes of the virtual networks (e.g. traffic demand changes) or the substrate network (e.g. network resource cost changes), embedding solutions might become sub-optimal over time. By enabling multi-step migration processes, our model takes advantage of so-called soft bottlenecks. Soft bottlenecks refer to capacity bottlenecks in the substrate network imposed by other virtual network demands which can be mitigated by releasing some of these existing demands when multiple migration steps are involved. Our model is separated into two phases. Phase I determines the optimum embedding solution, based on the current cost values, while Phase II targets the optimum migration process i.e. the transition towards the optimum embedding solution applying the concept of Time-Expanded Networks [4]. The evaluation of our model is performed using a realistic substrate network topology from SNDlib [12].