Research On Privacy-Preserving Deployment And Migration For Service Function Chains Across Domains Based On Deep Reinforcement Learning

The capacity of network function virtualization(NFV)to dynamically define and manage network functions is drawing attention.An ordered connected collection of virtual network functions(VNF)that are defined using NFVS is known as a service function chain(SFC).Flexible customization and provision of network services are possible.A single infrastructure provider cannot enable the deployment of all VNFs and fulfill all network functions inside the complete physical network under the large-scale,heterogeneous Industrial Internet of Things(IIoT),therefore cross-domain deployment of SFC is necessary.Cross-domain deployment of SFC,however,invariably results in privacy leakage and problems with resource coordination among several domains.In order to solve the problems of privacy protection and cross-domain resource coordination during cross-domain deployment of SFC and increase the effectiveness of network services,this paper proposes a cross-domain deployment mechanism of SFC privacy protection based on deep reinforcement learning(DRL)and an optimization mechanism of SFC deployment based on VNF migration optimization algorithm.Here are the specifics:(1)It is suggested that a resource prediction and binary response approach be used to address the issue of privacy leaking across various domains.This method involves the multi-domain controller sending the intra-domain controller the user’s SFC request,which is then used to predict the virtual resources needed by each node by comparing the resources needed by each VNF with the resources available in each domain.The service intent response matrix(SIRM)is created concurrently without publishing any node data to the domain.The SFC deployment is then finished using SIRM as input,which is based on the deep Q network.The suggested approach is contrasted with the resource-exposing ERI-SP and ERI-DQN algorithms as well as the column-generation privacy-protecting CC-BSFC technique.The experimental findings demonstrate that the suggested solution may decrease the end-to-end latency by up to 19ms and increase the SFC acceptance rate by more than 90%while maintaining anonymity.(2)A suggested optimization approach for SFC deployment is based on the VNF migration optimization algorithm and targets the dynamic change of network traffic and aberrant mutation of network nodes.A novel migration strategy is developed based on the current SFC deployment decisions by separating the three network states of stable network,dynamic changing network,and node failure.Simulation tests are then conducted under each of the three network states.According to experimental findings,the suggested strategy may increase SFC’s resource consumption rate by 70-100%,boost deployment success for SFC to 52-64%in the event of a traffic spike or node failure,and make sure the system’s end-to-end latency is within acceptable bounds.The technological underpinnings for the cross-domain deployment of SFC in IIoT scenarios,intra-domain privacy protection,and VNF migration optimization will be presented in this article.These technical contributions will have significant practical implications for future research into the deployment strategies of SFC in multi-domain scenarios.