Breaking the atomicity of virtual network embedding

Network virtualization currently becomes an important technology in optimizing resource management in a datacenter. The major research issue in this area is the virtual network embedding problem (VNEP) concerned with mapping virtual networks (VNs) onto a shared substrate network (SN) with some constrains satisfied. In this paper, by proposing the new concepts of “node sharing” and “partially accepting”, we try to break the “atomicity” of the traditional solutions of VNEP, which usually treat a VN request as an inseparable whole. We consider a more realistic scenario where virtual nodes in a VN request are of different importance and accordingly devise a mapping algorithm called Best Effort Algorithm (BEA). Our algorithm can make the most of the resources through node sharing at idle period of the substrate network and provide the best-effort mapping by partially accepting VN requests when physical resources become scarce. Final simulations demonstrate the effectiveness of our algorithm and show that our novel ideas can lead to higher acceptance ratio and long-term average revenue.