Cost-Efficient Virtual Network Function Graph (vNFG) Provisioning in Multidomain Elastic Optical Networks

Abstract

Acting as a promising technology to bring revolutionary changes to how the networks are architected, network function virtualization (NFV) leverages IT virtualization technologies to instantiate various types of virtual network functions (vNFs) flexibly and dynamically on commodity hardware, which can be easily found in a variety of NFV infrastructure points-of-presence, e.g., datacenters. In this paper, we investigate how to cost-effectively provision vNF graphs (vNFGs) with arbitrary topologies in a multidomain elastic optical network that consists of two domains, i.e., the private and public ones. We first formulate an integer linear programming (ILP) model with the objective to minimize the total resource cost of vNFG provisioning, and show that it can solve the problem exactly. With the ILP model, we analyze the complexity of the problem and prove that it is an $\mathcal {NP}$ -hard one. Then, we leverage the minimum $k$ -cut problem to design two time-efficient heuristics. The results from extensive simulations verify the performance of our proposed algorithms, and indicate that they can balance the IT and spectrum resource usages intelligently according to the resource price setting in multidomain environment.