Research On The Resource Management For Network Virtualization And Virtual Network Application

Network virtualization is viewed as a promising way for overcoming the ossification of the Internet, which gets more and more researcher’s attention in recent years. Network virtualization intends to construct a completely virtualized network environment, which allowed multiple heterogeneous and diverse virtual networks to coexist in a shared physical substrate network by abstracting, isolating and re-building the public network infrastructure. Novel network protocols and applications can be deployed in one virtual network and will not influence data service deployed in other virtual networks. Virtual network can totally control the allocated virtual resource. Namely, virtual network can construct the network structure and protocols according to the transaction and the user demand, and also adjust the node resource and the link resource according to the dynamic environment. As the key technology of constructing the next generation Internet, network virtualization make the implementation and application of virtual network more flexible, but bring great challenge to the network resource management. In addition, network virtualization provides us a new view to solve the problem of the existing network, how to implement the network virtualization and combine the network virtualization with the existing network environment is also the research emphasis and difficulties.This paper focuses on the resource management in network virtualization environment and the routing optimization technology based on network virtualization technology, dedicates to improve the performance of network resource and sovle the routing problem of traditional network using network virtualization. To the resource management problem in network virtualization environment, game theories are introduced to expand research domain. This paper considers the revenue of the substrate network and the virtual network. With regard to different application scenarios, we proposed two resource management models based on different game theory method. To the application of network virtualization technology, this paper focuses on the failure recovery problem in IP layer and the multicast problem of the exsiting Internet environment. A fast re-route scheme based on network virtualization technology and a centralized multicast mechanism are presented. Prototype experiment and simulation experiment verified the feasibility and performance of the mechanism. Achieved theory results and experiment results may have certain reference value and practical meaning for the network virtualization research.Specifically, the key contributions in the dissertation are listed as follows:(1) In network virtualization environment, interest involves among substrate networks, virtual networks and the end users. Maximizing the revenue of substrate network and solving the resource demand problem of virtual network simultaneously is an important issue. To solve the resource allocation problem in network virtualization environment, this paper proposes a pricing decision and resource allocation model to maximize the revenue of substrate network and virtual network simultaneously based on Stackelberg game. In this model, the revenue function of the substrate network and the virtual network are introduced. Then, the optimal strategies of both players in complete information state are given, and a transfer payment scheme is proposed which will get more revenue for both players in cooperation decision model. Whereas, in the incomplete information state, the existence of Nash equilibrium point of the non-cooperative game among the virtual networks is proved. To obtain the optimal strategies of virtual network bandwidth allocation and physical network pricing, a distribution iteration algorithm was presented. Finally, the simulation results demonstrated the effectiveness of the algorithm, which can reach the equilibrium and ensure the stability of network in the optimal state.(2) Viewed from the virtual network point, this paper models the substrate network path selection problem of virtual network as an evolutionary game. The virtual networks used the same substrate path are abstracted as a population. The virtual networks in the population choose a best substrate path by the repeatedly study. In the process of evolutionary game, virtual networks adjust the strategies through the iterated game. This paper leverages the replicator dynamics equation to analyze the dynamic variation of strategies selection in the population and demonstrates that this dynamics equation is the admissible dynamic to the evolutionary game. This paper also uses the potential game theory to analyze the evolutionary process and demonstrates the evolutionary result will converge to the evolutionary equilibrium. At the evolutionary equilibrium, all the virtual networks in the populations can get the best selection strategies. Finally, an evolutionary algorithm is proposed. The experiments simulated the process of virtual network selecting the substrate paths in different initial state and reached the evolutionary equilibrium. Simulation results are coincided with the theoretical analysis results. (3) By studying the characteristics of IP layer failure model and existing IP layer network failure recovery technologies, this paper proposes a fast re-route scheme based on OpenFlow technology. The proposed scheme viewed the OpenFlow network as the substrate network which can forward data arbitrarily, and viewed the IP network as one of the virtual networks of the substrate network. The core idea of this scheme is transferring the IP layer failure problem to the substrate layer network. We exploit OpenFlow technology and deploy a fast failure detection model into the existing routers. The routing information database (RIB) and the flow table construct a two-layer forwarding structure. The proposed scheme contain three panels:control panel, routing forwarding panel and OpenFlow forwarding panel.When the link failure happens, fast failure detection model notify the control panel. Control node computed the backup paths for the destinations that are affected by the failure link and sent the related flow table to the forwarders.During the interval of link failure, the OpenFlow forwarding panel will forward influenced data to the destination which will lower the packet loss during the convergence of routing protocol. Finally, this paper designed a prototype system to verify the performance of this scheme. Experiments results demonstrated that the proposed framework have a low packet loss to the link failure.(4) This paper proposes a hierarchical, forward node clustered management and centralized multicast mechanism, which separate the data plane and control plane by shifting the multicast management to specific deployed servers. Middle nodes forward data based on the rule sent by the servers. The servers constructed a Full-Mesh topology, and the clustering and centralized management of forward node efficiently improve the multicast forwarding ability. The protocol between the controllers and the constructing and maintenance algorithm of multicast tree are introduced. At last, a simulation experiment and a prototype experiment are presented to verify the proposed mechanism. Compared with traditional multicast, the proposed mechanism can alleviate the burden of multicast routers and also accelerate the router’s packet forwarding speed. Simulation and experiment results demonstrate that OFM has a good performance. Compared with the multicast data, the proposed mechanism has a smaller protocol overhead.