Research On Embedding Algorithm For Network Virtualization

The rigid problem of current network architecture becomes more and more serious, making it difficult to provide differentiated network services. Although VPN, Overlay network and other technologies have been proposed to solve the problem, the application of those technologies is very limited. It's the Virtualization technology that makes cloud computing achieve an important means to maximize the use of computing resources, but the existing resource allocation solutions only provide virtual machine instances, because of the lack of effective management of network resources, resulting that some applications sensitive to network resources is hard to obtain a stable network resources and then run in higher operating efficiency.Network virtualization technology is a network abstraction technology to address the rigidities of existing networks by embedding multiple independent virtual networks with flexible topologies into the same physical network. The important problem of network virtualization technology to be solved is the embedding of the virtual network, and the essence of the embedding problem is to use the physical network resource as efficient as possible under the condition that reliable network service should be guaranteed. This paper focuses on the improvement of the virtual network embedding algorithm used to solve the embedding problem, and the main work includes the following three aspects:1. This paper briefly summarizes the development background and significance of the network virtualization, and presents the model of the virtual network embedding problem related to network virtualization, simultaneously listing several performance indicators for evaluating different virtual network embedding algorithm. Moreover, a comprehensive summary of the existing virtual network embedding algorithm is made, and several aspects are considered to classify those algorithm. Lastly, We analyze several classical embedding algorithm based on their application environment.2. In order to solve the problem that most of existing virtual network embedding algorithms don't consider the resource consumption of intermediate hidden nodes, this paper uses the adjacent nodes in the virtual network as physical network resource allocation unit, and the more the potential data traffic between them, the more likely to get get the best physical network resources. Further, the paper presents the concept of the intimacy between two adjacent nodes, which is used to measure the potential data traffic, and demonstrates the efficiency of the embedding algorithm based on the intimacy from the theoretical and exmperimental points of view.3. Aiming at the problem that some existing virtual network embedding alogritm are difficult to be applied to large-scale virtual network embedding proble, the paper proposes using the improved multilevel recursive bisection method to divide the original virtual network into multiple small sub-virtual networks with inter-associations as little as possible, and then uses the existing embedding algorithm to map. Further, the paper demonstrates that the proposed computation method for virtual network partition cost is more suitable for virtual network partition problem from the theoretical and exmperimental points of view.