Subswarm-guided ant colony optimization with enhanced pheromone update mechanism and beam search for VNF placement and routing

Abstract

Network Function Virtualization (NFV) is a novel technology that enables flexible and cost-effective networks by replacing the hardware-based middleboxes with software run on virtual machines called virtual network function (VNF). In NFV, each Service Function Chain (SFC) is required to be a set of ordered VNF from that must be optimally located across servers/distributed data centers. This paper studies the multi-objective virtual network function placement and routing problem (MO-VNFPR), which involves placing VNFs optimally in servers nodes and assigning server resources efficiently to these VNFs to satisfy user’s requests in the networks. Most existing methods for this problem only handled one requests or considered VNF placement and SFC routing separately. In this paper, we consider MO-VNFPR with two objectives: (i) minimize the overall cost by deploying servers and installing VNFs; (ii) reduce the network delay. A metaheuristic called subswarm-guided ant colony optimization (named SgACO) is introduced to efficiently solve MO-VNFPR. In SgACO, a subswarm-guided mechanism is adopted to satisfy requests at once by reorganizing the ant colony into subswarms. Moreover, an enhanced pheromone update mechanism and beam search are developed to investigate the search space efficiently. Experimental simulations on COGENT, CONUS, and NSF network topologies demonstrate that our algorithm achieves competitive results and performs better against other existing algorithms.