Abstract:
In this study, our objective was to address the drawback of the inability to effectively deal with hardware failure in existing virtual network embedding methods for cloud services. We propose a highly reliable virtual-network embedding strategy based on cooperative game theory (CG-VNE). CG-VNE maximizes cloud provider revenue by maximizing the client acceptance rate, as well as minimizing the blackout rate of virtual networks caused by the outage of substrate routers and/or links. To skirt the exponential complexity of virtual network mapping, in CG-VNE, the virtual network embedding problem is described as two interlocking and cooperative games. The first games deal with the virtual node mapping problem, and the second deal with the embedding problem of the virtual link. Through these two games, the virtual game can cooperate to achieve Nash equilibrium. In addition, the physical failures of the router and link are effectively handled while increasing cloud provider revenue. The results of our comprehensive simulation experiments show that, compared with the current majority of virtual network embedding algorithms, CG-VNE significantly reduces the rejection rate of new clients, increases cloud provider revenue, and lowers the rate of clients impacted by physical failures.