Study On Service Function Chain Deployment And Backup Methods Under NFV Environment

As user demand for services increases exponentially,traditional networks that rely on dedicated network hardware devices to implement network functions have failed to meet user needs.Service providers migrate network functions from dedicated hardware devices to software middle-boxes through network function virtualization technology to flexibly provide users with various functional services.In network function virtualization,the service function chain has become a popular form of network service.The reasonable allocation of network resources to the service function chain has gradually become an important research direction for people to study network problems.This paper studies the deployment and backup problems in the resource allocation of the service function chain.The main research contents are as follows:1)Aiming at the problem of weak adaptive ability of service function chain online deployment,this paper studies an online deployment algorithm of service function chain based on deep reinforcement learning.With the goal of maximizing long-term average revenue,this paper formulates the service function chain deployment problem as a combinatorial optimization constraint problem.In this paper,a node sorting algorithm generates multiple candidate actions in a probabilistic manner to reduce the time complexity.An efficient function approximation method based on graph convolutional neural networks is proposed to predict the future long-term revenue potential in the network.In addition,we employ a temporal-difference learning algorithm to make online deployment decisions for service function chain requests arriving in sequence.Finally,a prioritized experience replay technique is used to speed up the parameter update of the neural network.Experimental results demonstrate that our approach improves 10.8%,11.2% and 18.7% compared with the existing algorithms in terms of long-term average revenue,acceptance ratio,and revenue-to-cost ratio.In addition,our approach also has good robustness under different working conditions.2)The unreliability of software and hardware in the network will reduce the user experience.How to ensure that the service reliability and service level agreement will not be affected,so as to provide users with high-quality services is an important research direction.Therefore,this paper studies a reliability-based service function chain backup algorithm.Aiming at the problem of the backup model and state idealization of the service function chain,this paper considers the simultaneous backup of software and hardware while satisfying the reliability of the service function chain.In the state of resource heterogeneity,with the goal of minimizing backup resource consumption,the problem of service function chain backup is modeled as a combinatorial optimization constraint problem.Based on the two states of non-consolidation and consolidation,this paper studies two heuristic-based service function chain backup algorithms.In particular,in the consolidation state,this paper studies a heuristic-based efficient backup algorithm that merges adjacent virtual network functions as much as possible to reduce link backup resources.Simulation results show that our backup approach is superior to other existing algorithms in backup resource overhead and average acceptance rate by 51.7% and 44.3%.