Research On Virtual Network Embedded Model And Algorithm Based On Graph Entropy

The rapid development of different type demands for Internet applications increases the demand for Internet underlying hardware facilities networks. As users cannot afford the high cost of hardware facilities, Virtual Network(VN) service provider abstracts Infrastructure network and resources network user requests as Substrate Network(SN) and VN, respectively, and uses the network virtualization(NV) technology to make network resources as paid service to users. As a key part of network virtualization technology, Virtual Network Embedding(VNE) has been focused. The researches of traditional VNE algorithm improve VN request acceptance ratio but ignore the needs form users. This thesis study VNE problem, and focused on VN request acceptance ratio as while as Quality of Service(QoS).Chapter 1 of this thesis describes the background and the significance of research VNE issues. Then, the research status of VNE problem and the main contents of this article are presented.Chapter 2 presents VNE issues in detail. Firstly, for better understanding VNE process, the general model of VNE process is expounded. Secondly, the VNE business models help us understand the Internet future architecture and realize the effect of VNE algorithm on commercial members. The lastly, some common parameters used to evaluate the performance of the VNE algorithm are introduced, as well the definitions of those performance parameters.In Chapter 3, the author proposed and studied the issue of reducing the stress for substrate links in VNE process. When some VN requests with pressure-sensitive use some existing VNE algorithms, they can not guarantee QoS for the user. Then the author proposed a link stress-aware VNE algorithm(LSAVNE). And this research considered network resource properties of each node as its service priority. In order to shorten substrate path for virtual link, the author made the area whose center is the first virtual node embed in and whose radius is determined by VN characters as VN embed scope. On the stage of link embedding, the author found the suitable path for the virtual link in the k shortest paths with the pre-requested resource method. In the comparison experiments with deterministic node embedding with k-shortest path link embedding algorithm/splittable link embedding algorithms, LSAVNE has better performance on costs, cost / revenue, average link stress, VNR acceptance ratio than the others.Chapter 4 found traditional algorithms ignore VN candidate objects have different performances for VN and SN. When VN chooses its similar structure to embed in, it has better performance than random selective structure. So the author considered the graph entropy which can quantize the information structure in VN and SN and can judge similarity between candidate objects and VN. As complex calculation of graph-cut methods, a resources search method was proposed to avoid this issue. For the above problems,the author modeled these problem and proposed a virtual network embedding algorithm based on entropy(BEVNE) in this study. And on the simulation platform, simulation experiments were proceeded to verify the correctness and validity of the model and the algorithm.Chapter 5 summarized the research and gave direction for the future.