Research On Fault Diagnosis Algorithm In The Network Virtualization Environment

With the rapid development of modern communication technology, the traditional Internet is facing more and more challenges. It is increasingly difficult for the traditional Internet to adapt to a variety of new network services and to meet the huge needs of users. Network virtualization technology which is an important component of the future network architecture is one of effective ways to solve the existing impasse of the Internet. The mapping between SN and VN increases the complexity of fault propagation model, and some technologies such as node migration and link segmentation also increase the probability that VN service fault occurs. The rapid increase in the network service demands causes the large-scale set of faults and symptoms, which leads to the higher complexity of fault diagnosis. In addition, the operational state of the equipment in the substrate network is unknown to network service providers. Therefore, the research on fault diagnosis algorithm in network virtualization environment has the important theoretical and practical significance.The faults in network virtualization environment are analyzed in terms of uncertainty and relevance in the paper. The fault diagnosis algorithm in network virtualization environment is researched profoundly based on related research results about network mapping. The main work is as follows:The basic concept of mapping mechanism is introduced in the paper. Node mapping, link mapping and mapping constraint functions are analyzed in detail. The paper adopts a two-stage heuristic algorithm to realize the virtual network mapping and simulates network virtualization environment.The layered fault propagation model in network virtualization environment is constructed in this paper. The paper proposes the probability-weighted Bayesian network model with detailed analysis of the property and application of probabilistic graphical models. The fault propagation model is constructed based on the relationship between faults and symptoms. According to the relationship of observed symptoms, possible faults and components in the substrate network, the layered fault propagation model is simplified, which reduces the complexity of the fault propagation model.The fault diagnosis algorithm based on dependency and suspected degree is proposed in this paper. The fault diagnosis algorithm based on dependency solves the problem of the extensive set of faults and the fault diagnosis algorithm based on suspected degree solves the problem of the extensive set of symptoms. The simulation results show that the fault diagnosis algorithm based on dependency and suspected degree acquires the higher diagnosis rate and the lower false positive rate in different experiment scenarios.