The Technologies Of Network Resource Allocation And Fault Diagnosis For Network Virtualization Environment

Network virtualization has been identified as a feasible approach to solve the problems of the Internet, which has caused attention of more and more researchers. Under this background, this dissertation focuses on the five problems of network resource allocation and fault management in the network virtualization environment. There are three problems about existing researches of resource allocation for the network virtualization environment, which include:(1) among competitive environment of multiple Infrastructure Providers (InPs) and multiple Service Providers (SPs), algorithms of resource allocation are ineffective, and trading market is not fair,(2) within the large scale substrate network environment, the existing mapping algorithms is inefficiency,(3) reconfiguration time is unreasonable, which results in some negative effect on network performance. At the same time, there are two problems about existing researches of fault diagnosis for the network virtualization environment, which include:(1) SP can not diagnose all service faults of virtual network, because the information of substrate network is invisible for SP,(2) the false positive rate is higher and the running time is longer because of the large fault set and the large symptom set, which is caused by each substrate network bearing multiple virtual networks.According to the problems mentioned above, on the basis of the existing researches of resource allocation and fault diagnosis for the network virtualization environment, and benefiting from technologies related to the current network management, the main contributions of this dissertation are as follows: (1) To solve the problems among competitive environment of multiple InPs and SPs, that is, algorithms of resource allocation are ineffective, and trading market is not fair, auction based virtual network resource allocation mechanism (ABVNRAM) is presented from service management aspect. Experiments show that ABVNRAM can complete multiple InPs and SPs resource allocation and pricing each time. The fairest trading market is established by ABVNRAM than V-MART and ABVNRAM without bargaining.(2) In order to improve the efficiency of the existing mapping algorithms, Virtual Network Mapping Algorithm in the Shortest Mapping Time (VNMAiSMT) is presented. VNMAiSMT includes community partitioning algorithm based on k-means clustering and resource allocation algorithm. Experiments show that VNMAiSMT algorithm, compared with D-ViNE, reduces the mapping time and improves the mapping efficiency.(3) To reduce the negative effect of reconfiguration on network performance and calculate the appropriate reconfiguration time, mathematic model of the utilization rate of substrate network and reconfiguration time is established, the calculating method of reconfiguration time is designed, and the limit number of reconfiguration requests was deduced. Secondly, forecast-based resource reconfiguration algorithm (FRRA) for network virtualization environment is proposed. The results show that algorithm FRRA reduces reconfiguration cost, improves virtual network request reception rate.(4) To solve the difficult problems of fault diagnosis caused by the invisible of the information of substrate network for SP, service fault propagation model based on mapping relationship is presented, based on the mapping relationship between virtual network and substrate network. To enhance the performance of fault diagnosis algorithms, service fault diagnosis algorithm based on inherent correlation among symptoms (SFDoIC) is proposed. Simulation results show that algorithm SFDoIC can solve the difficult problems of fault diagnosis caused by the invisible of the information of substrate network for SP, effectively reduce the false positive rate and decrease running time.