Research On High Concurrency Services Recovery Mechanism For F5G

As an important cornerstone of the development of the Internet,communication network has made a significant contribution to the transformation of social informatization.At present,the complex and diversified service types and service requirements of the Internet have higher requirements for network quality,bandwidth and delay.With the development of the fifth generation fixed network,optical network sinks to the edge,which increases the scale of the network and the number of services it carries,bringing great pressure to the management and control performance of the network.When an optical fiber in a network is faulty due to human or geographic factors,there will be thousands of services interrupted and network quality is seriously affected.Faced with the recovery request of high concurrent services in large-scale networks,the traditional recovery mechanism cannot meet the service requirements in terms of recovery effect and recovery delay.Therefore,this paper focuses on the recovery of highly concurrent services in large-scale networks,designs a centralized and distributed network architecture for F5G,and optimizes the route calculation and signaling delivery in the process of service recovery.The goal is to reduce the path calculation experiment in the recovery process to achieve fast path calculation of high concurrent services,reduce the signaling delivery delay in the recovery process to achieve fast signaling delivery,and coordinate the route calculation and signaling delivery to achieve fast link recovery.In this paper,we first propose a centralized and distributed network architecture for F5G,and then propose three innovative solutions to the problem of high concurrency service recovery.Specific innovation work is as follows:First,in view of the problem that traditional routing algorithms consume too much time delay in the face of high concurrency routing requests,we propose a fast routing mechanism for high concurrency services based on deep learning prediction.Abstract a large-scale network into a small-scale network through topology abstraction,calculate a virtual path for services in the small-scale network,and learn the mapping relationship between the virtual path and the real path through the deep learning model,so as to realize the function of path calculation.The simulation results show that the model can reduce the path calculation delay in large-scale networks,but the accuracy of the model decreases with the increase of network scale.When the network node size is 3015,the accuracy of the model can reach 87.17%.Meanwhile,compared with Dijkstra algorithm,the proposed scheme reduce the path calculation delay by 2.16msSecond,we propose three signaling delivery mechanisms based on the combination of centralized and distributed signaling to solve the problem of high signaling delay and high signaling overhead in the traditional signaling delivery mechanism.They are the high concurrency signaling delivery mechanism based on centralized triggering centralized triggering distributed centralized triggering high concurrency signaling delivery mechanism based on path segmentation.The simulation results show that the high concurrency signaling fast delivery mechanism based on path segmentation is the best in the thousand-level network topology.Compared with the traditional centralized scheme,the signaling cost increases by 23.5%,but the signaling delay decreases by 45.6%.Compared with the traditional distributed scheme,the signaling delay increase by 4.2%,but reduce the signaling cost by 47.5%.Thirdly,in view of the low recovery success rate and large recovery delay of traditional link recovery mechanism in the face of high concurrency service recovery requests,a high concurrency service fast link recovery mechanism based on conflict avoidance is proposed.By increasing the range of link recovery,there will be more network resources using to achieve a higher recovery success rate.In addition,when signaling is delivered,the signaling information of different services on the same recovery path is combined to reduce the number of signaling that needs to be delivered,thus reducing the signaling delivery delay and realizing fast link recovery for highly concurrent services.The simulation results show that the scheme can improve the recovery success rate by 11%compared with the traditional KSP algorithm.At the same time,the proposed scheme can effectively reduce the recovery delay in large-scale networks.