Optimal Traffic Engineering and Placement of Virtual Machines in SDNs with Service Chaining

Abstract

In this paper, we investigate the service capacity of a software-defined network with network function virtualization and service chaining. To obtain and achieve the service capacity, we formulate a mixed integer linear programming problem to jointly optimize the placement of virtual machines and the steering of network flows. In addition, we propose the greedy linear programming relaxation algorithm for efficiently obtaining solutions in three phases based on auxiliary linear programming problems and augmented service paths. We justify the usage of the proposed approach by analytical results and simulation results.