Optimal VNF Placement and Traffic Steering using Column Generation

Abstract

VNF placement and traffic steering have received considerable attention in the literature, ever since business pressures started driving telecom operators to embrace virtualization technologies instead of continuing with the expensive “dedicated hardware and appliance” approach. However, most proposed solutions are not exact. When the formulated problem results in an Integer Linear Program (ILP) or a Mixed Integer Linear Program (MILP), the focus shifts to searching for heuristics. We show that using the Column Generation (CG) technique, exact solutions can be obtained quickly. The CG approach can be used to solve problems in both offline and online scenarios. For the aggregate bandwidth minimization problem, the exact solution provided by CG improves significantly upon “DP-COA”, a recently proposed heuristic. However, aggregate bandwidth minimization can lead to highly uneven link loading. Motivated by this, we formulate another problem in which the objective is to minimize the maximum link bandwidth consumed over all links in the network. This problem is also solved exactly using CG. Results indicate that this approach can help achieve balanced traffic demand placement.