Research On Fault Detection And Healing Algorithms In Network Slicing

The application of Network Function Virtulization(NFV)and Software Defined Network(SDN)technology enables network slicing to achieve software and hardware decoupling,thereby providing fast,flexible,and diversified services.While meeting future business needs,network slicing also introduces additional complexity and loses the high reliability guarantee provided by dedicated equipment,which brings greater challenges to network service guarantee and fault management.In order to solve the problem of fault management in network slicing,it is critical to carry out the corresponding self-healing technology research,that is,it is necessary to automatically detect,diagnose and recover the faults in network slicing.This article focuses on fault detection and healing in network slicing scenarios.The work content is as follows:1.In order to quickly detect virtual network function(VNF)faults and solve the communication overhead and security privacy issues in centralized VNF fault detection,this paper proposes a CNN-GRU fault detection model based on federated learning.Accurately detect VNF failures.First,based on the isolation characteristics of network slicing and the characteristic that the Service Function Chain(SFC)is composed of a set of VNFs in an orderly manner,a federated learning framework for VNF failure detection in network slicing scenarios is established.Secondly,an unsupervised fault detection model based on CNN-GRU is proposed.The CNN-GRU model is used to predict the VNF performance value,and then the abnormal score is used to detect whether the VNF is faulty.Finally,a gradient compression mechanism and an adaptive optimization method are used in federated learning to reduce the communication overhead in federated learning and enhance the convergence effect of federated learning.The simulation results show that the model can ensure the effect of VNF fault detection while protecting data security and privacy,reduce the communication overhead of federated learning,and optimize the convergence effect of federated learning.2.In order to realize the rapid healing of faulty VNFs and ensure the network's service quality and high reliability requirements,this paper proposes a faulty VNF healing algorithm based on deep reinforcement learning.First,the potential reliability of the physical server is defined according to the state of the VNF to assist in the VNF healing decision.Secondly,a VNF healing cost model is established based on healing time,network load balancing and network potential reliability to comprehensively consider the VNF healing time and network reliability requirements.Finally,the Markov Decision Process(MDP)is used to express the healing problem of the faulty VNF,and deep reinforcement learning is used to make the decision of the VNF healing strategy.The simulation results show that the algorithm can effectively reduce the healing time of faulty VNFs,while ensuring the load balance and reliability requirements of the network.