An Efficient VNF Deployment Scheme for Cloud Networks

With maturity of cloud network technologies and rapid development of NFV (Network Function Virtualization), the research on NFV based on cloud technologies has gradually increased in academics and industry, aiming to improve scalability, flexibility, and cost-efficiency of network service provision systems. A key challenge of NFV is how to design a cost-effective VNF (Virtualized Network Function) deployment scheme. Most of the existing schemes deploy VNF based on resources requested by users and ignore optimizations during the running time, which causes low capacity utilization. Therefore, we propose an efficient VNF deployment scheme in this paper, which has two parts: over-deployment and automatic migration to improve resource utilization rate and guarantee on-demand capacity provisioning at the same time. The over-deployment approach takes advantage of the characteristic that resources actually used by VNFs are mostly less than the resources requested and deploys VNF based on running load. However, over-deployment may cause PMs (Physical Machines) overloaded at some time, thus, we use the automatic migration approach to solve the overload problem by migrating VNFs. The automatic migration approach moves proper VNFs to lighten the load of overloaded PMs. We perform extensive simulations and numerical results verify benefits of the proposed scheme in terms of the average number of deployed VNFs, active PMs, the rejection rate of VNF requests, and VNF migrations.