Research On Virtual Optical Network Mapping Algorithm

With the emergence of new services such as data centers,the architecture of the existing network has not adapted to the development of new services.Through cost-effective technologies such as network virtualization,operators abstract physical networks into virtual resources and allocate them into multiple independent virtual networks to share physical resources and deploy services according to the diverse needs of users.Today's networks require dynamic loading and network resources on demand.The emergence of Network Function Virtualization(NFV)has become an effective way to solve the problem of network rigidity.Elastic optical networks have more granular spectrum resources to achieve more flexible spectrum bandwidth allocation and provide the spectrum resources closest to the demand according to service requests,thereby reducing resource over-allocation and improving spectrum utilization.Therefore,in the future,the elastic optical network can become the bearer network of the NFV,so that it can meet the various business needs of the operators without changing the current network architecture.Multi-point to multi-point services such as new data center cloud computing can be abstracted into virtual optical network services.Therefore,after network virtualization is implemented on the elastic optical network,operators need to allocate reasonable resources for virtual optical network services in order to increase revenue and provide better services,and also need to consider cost and efficiency.This is the research significance of virtual optical network resource mapping algorithm research.The main research results of this paper are as follows:(1)For the problem of virtual node priority mapping in virtual optical network mapping,this paper proposes a new virtual node priority mapping algorithm.Based on the distance-adaptive node-first mapping algorithm,the algorithm adds analysis of current virtual link bandwidth requirements and link spectrum usage during the link mapping phase.By introducing an auxiliary map based on the virtual link bandwidth request,the non-compliant physical link is pruned,a set of candidate paths with high spectral continuity is selected,and the shortest path is selected according to the distance sorting.The simulation shows that compared with the existing node-first mapping algorithm based on maximum computing resources and the distance-adaptive node-first mapping algorithm,the new algorithm reduces the blocking rate by about 30%.(2)For the problem of virtual link priority mapping in virtual optical network mapping,this paper proposes a new resource weighted sorting algorithm.When the virtual link and the candidate path are sorted,the new algorithm weights the relative sizes of the node resources and the link resources to determine the order of the virtual link mapping and the selection priority of the candidate paths.Compared with the previous sorting algorithm using resource size,the new algorithm is more stable in order,more interpretable,and better in blocking rate performance.(3)For the scheduling problem of candidate paths in virtual link priority mapping,this paper proposes a new candidate path sorting algorithm.The new algorithm transforms the sorting problem of path scoring into a machine learning two-class problem,that is,the algorithm learns the probability of success of the virtual link mapping to the current need to evaluate the path.All candidate paths are sorted according to probability,and the path with the highest probability of success is selected.The simulation shows that the new algorithm has better blocking rate performance than the virtual link priority mapping algorithm based on the shortest path,and can be combined as a separate module with the resource weighted sorting algorithm in(2)to achieve better results.