Research On Network Slice Fault Recovery Under Cloud Network Integration

In recent years,cloud network integration has become the most important development trend in the information and communication industry.With the continuous popularization of cloud services,the position and role of the network in cloud services also need to be redefined.The network no longer only provides connection channels,but is more reflected as an important means of differentiated competition in cloud services.In this new context,how to ensure that a network slice that includes both cloud and network resources(the network slice discussed in this article is an end-to-end business that includes cloud computing resources and network transmission and forwarding resources)can quickly recover after a failure is the key to achieving stable network operation.This paper conducts research on this issue,and the main work content and innovative points are as follows:Firstly,to address the issues of low probability and uncertain generation time of fault data in fault detection,this paper proposes an unsupervised network slicing fault detection mechanism.This mechanism is based on variational autoencoder(VAE)and utilizes unsupervised means to train with only normal data.The simulation results show that the accuracy of the fault detection mechanism can reach 99.00%after the model is stable,with a standard deviation of 1.05.This effectively overcomes the problem of collecting fault data while ensuring the accuracy of system recognition.Secondly,under the guarantee of real-time and effective fault detection mechanisms,a resource sufficient network slicing fault fast recovery algorithm is proposed for network slicing fault recovery in the context of cloud network fusion.This algorithm fully considers the mutual cooperation and constraints of multidimensional resources such as computation,storage,and transmission.In the recovery process,the selection of recovery nodes is no longer based on a single level factor,but rather on the internal relationship between cloud resources and network resources,comprehensively considering the impact of various factors on the network recovery process.The simulation results show that the proposed algorithm has a 5.3%and 21.8%improvement in recovery efficiency compared to traditional greedy recovery algorithms and random recovery algorithms,respectively.Finally,in response to the problems of limited and variable recovery resources in practical applications,this paper further proposes a fast recovery algorithm for multi resource aware network slices based on resource scarcity.This algorithm takes into account the limited resources and stage variability of the cloud network,taking into account the impact of different resources on the network recovery process.The simulation results show that for the two algorithms using Greedy as the base algorithm,the addition of resource aware algorithm improved the performance by 9.2%,and for the two algorithms using JSFR as the base algorithm,the addition of resource aware algorithm improved the performance by 6.4%.