Virtual Network Embedding And Network Resource Allocation Based On Forecast Mechanism

Internet is one of the most important infrastructures in today’s society. It deeply influences and changes people’s work and life. However, as the rapid expansion of Internet size and new application, the existing Internet architecture has not met the demands of modern society. A variety of problems in the existing network architecture has gradually exposed, such as safety, scalability, quality of service and energy consumption. In order to solve the problem and bottleneck that network architecture confronts, academia put forwards the design idea of "clean-slate". Researchers hope to get rid of the bondage of existing network architecture, and design new generation network architecture.What the future network should be has not formed unified viewpoint in academia, but the future network supporting virtualization has been widely recognized. Network virtualization technology provides an effective way to solve the ossification of current Internet. On the one hand, network virtualization technology supports networks of different architectures to share substrate network. It provides a solution for the validation of different new technologies and protocols. On the other hand, the core idea of network virtualization technology is to abstract and isolate the underlying network resources, so as to realize the separation of control plane and data plane. This feature is becoming the key attribute of the future network.Virtual network embedding is a challenge in network virtualization. It aims at how multiple virtual network requests with resource constraints to share the substrate network resource, and it is a typical NP-hard problem. In this paper, considering the degree of convergence when mapping a virtual network onto the substrate network, we propose a new embedding algorithm based on topology mapping convergence-degree. Recently, Software-defined Network (SDN) using network virtualization technology provides a solution for the future study of new technology and application. As a typical representative of SDN, OpenFlow technology attracts great attention of academia and industry. Therefore, based on the research of virtual network embedding, this paper further studies SDN architecture based on OpenFlow and its resource allocation strategy.This paper’s main innovations are as follows:(1) The paper makes the most favorable processing to the subsequent link embedding stage in the node embedding stage. Defining the topology mapping convergence-degree, we realize a new algorithm based on topology mapping convergence-degree. The algorithm largely decreases the complexity of embedding problem and improves the network utilization efficiency.(2) The paper defines the balance rate of link load to reflect the load balance of substrate links, and improves it largely. To the best of our knowledge, there are few articles researching it.(3) Based on the feature of star topology and its importance, we further propose a star topology embedding algorithm, and further improve algorithm’s performance.(4) Analyzing the problems of SDN architecture based on OpenFlow, the paper proposes a kind of hierarchical SDN architecture, and shows the scalability and robustness of the hierarchical SDN architecture.(5) Based on the hierarchical SDN architecture, this paper proposes a resource allocation strategy with prediction mechanism through importing preprocessing module in traditional OpenFlow switch, and improves the intelligence of SDN resource allocation.