Multi-Objective Optimization-Based Virtual Network Embedding Algorithm for Software-Defined Networking

Abstract

To overcome the drawbacks of traditional Internet architectures, software-defined networking (SDN) technology has been proposed, which is expected to dramatically simplify network control processes and enable the convenient deployment of sophisticated network functions. To achieve highly efficient resource utilization in SDN and offer users with diverse service requirements, virtual network embedding (VNE), which maps various virtual network requests of users to a given substrate network, should be conducted. In this paper, we study the VNE problem in SDN where the substrate SDN switches and links may be subject to malicious attacks. We first propose a hierarchical virtualization-enabled SDN architecture based on which the VNE strategy can be designed. Then, stressing the importance of network load and reliability of the substrate network, we formulate the VNE problem of SDN as a multi-objective optimization problem which jointly minimizes network load and maximizes embedding reliability under the constraints of virtual network requirements and the resource characteristics of substrate network. As the formulated optimization problem is a complicated multi-objective optimization problem which cannot be solved conveniently, we apply the ideal point method. In particular, we first propose virtual node embedding sub-algorithm and virtual link embedding sub-algorithm to determine the locally optimal solution to the two subproblems, i.e., network load minimization subproblem and embedding reliability maximization subproblem. Then, examining the distance between the feasible solutions and the locally optimal solutions, we formulate a single-objective optimization problem and solve the problem to obtain the global VNE strategy by applying discrete particle swarm optimization (DPSO) algorithm. Numerical results demonstrate the effectiveness of the proposed algorithm.