Research On Virtual-Network-Function Deployment Algorithm For Reliability Guarantee In Multi-Access Edge Computing

Multi-access Edge Computing(MEC)is a novel network architecture that addresses the conflict between the growing number of compute-intensive applications and portable mobile Internet-of-Thing(Io T)devices with limited computation resources.With the development of MEC and Network Function Virtualization(NFV),the deployment problem of Virtual Network Function(VNF)in MEC has become a popular research direction.Although VNF can provide services with high software reliability,the tasks and VNF on the servers still face execution failures and deployment failures when the edge servers suffer from failures and malfunctions.Therefore,we deploy VNF redundantly on different edge servers which assures the service reliability requirements of tasks.However,the redundancy deployment will greatly occupy computing resources and bring greater costs,so we propose the Profit Maximization problem for Reliability Assurance(PMRA).To solve this problem,we first determine the number of VNF redundancy deployments based on the reliability of the edge server hardware and the task reliability requirements.Then,because of the tight coupling between task reliability and edge server hardware architecture,we consider the case of homogeneous and heterogeneous edge server hardware,and propose an efficient approximation algorithm and a fast heuristic algorithm,respectively.Those algorithms make decisions on VNF deployment locations under the scenario of limited computation resources of edge servers.The main work is as follows:1.We use a distributed redundancy deployment strategy to meet the reliability requirements of the task and construct a PMRA problem.The aim is to maximize the profit of network service providers while assuring task reliability.The problem is proved to be an NP-Hard problem.2.First,we consider the PMRA problem in homogeneous edge computing systems(Homo-PMRA),and propose a Local-Ratio-Based Algorithm(LRBA)by transforming the optimization problem into a number of identical subproblems,and the approximation of the algorithm is proven.Then,we also consider the PMRA problem in heterogeneous edge computing systems(Heter-PMRA),and a Benefit-cost Ratio Prioritized Algorithm(BRPA)is proposed.The algorithm prioritizes the deployment of VNF on highly reliable servers,and is able to find the approximate solution quickly.3.Finally,simulation experiments are carried for two network scenarios,where we compare LRBA and BRPA with local search algorithm and random algorithm in the homogeneous scenario,and compare BRPA with greedy algorithm and random algorithm in the heterogeneous scenario.It is proved that the proposed LRBA and BRPA have better performance than compared algorithms.