Research On Slice Admission Control And Embedding Optimization In SDN/NFV Networks

Network slicing is one of the core supporting technologies of 5G network,infrastructure network providers will create different network slicing instances based on the needs of different business scenarios.How to optimize the efficiency of different resource allocation processes during the life cycle of network slice instances is a hot topic in current research.A network slice instance will go through the following three resource allocation processes during its life cycle:(1)admission control;(2)slice embedding,also known as “slice deployment”;(3)slice reconfiguration.Among them: the admission control process is responsible for filtering out the set of network slice requests that meet specific optimization goals;the slice deployment process is responsible for selecting reasonable physical nodes and links to deploy slice requests;the slice reconfiguration process is responsible for re-configuring the amount of resources of slice instances that cannot meet user needs.Most of the known studies only consider the optimization of slice deployment,and take maximizing the success rate of slice deployment as the optimization goal.The problem with this kind of research is: according to the greedy strategy,the objective function is set to minimize the deployment cost of a single request.Although a high overall acceptance rate can be obtained according to the objective function under the premise of a small number of requests are deployed in a network with relatively sufficient resources,however,when the network size becomes larger and the number of requests becomes larger,the objective function will cause the underlying network to have resource fragmentation problems,and the acceptance rate will therefore decrease to a certain extent.In recent years,some studies have jointly optimized the processes of admission control and slice deployment.The problem with this kind of researches is that such researches only use maximizing the benefits of network services as the objective function for single-objective optimization.Although it can maximize the benefits obtained by infrastructure providers,ignoring the fairness between different service types will lead to poor user experience.This thesis studies the problems existing in the above two types of research,and the main work is as follows:(1)In view of the resource fragmentation problem that occurs when deploying a large number of requests on a large-scale network in the existing slice deployment research,this thesis firstly proposes a slice deployment method based on Generalized Network Temperature(Generalized Network Temperature)GNT-MCTS-VNE,which alleviates the problem of resource fragmentation;secondly,this thesis proposes a pruning strategy RLS＿NR based on node importance ranking and path length limiting,which reduces the time complexity of the algorithm under the premise of ensuring that the acceptance rate does not drop drastically.(2)In view of the problems of excessive resource consumption of high-priority slices and lack of fairness between services in the research on joint slice admission control and deployment optimization,this thesis firstly uses the Jain index to redesign the objective function of joint optimization.Secondly,a two-stage joint optimization algorithm WAC-NIE is proposed.The joint queue selection and discrete judgment admission control method of WAC-NIE effectively improves the overall acceptance rate.At the same time,WAC-NIE effectively balances the fairness between different services under the premise of ensuring that service revenue does not decline drastically.(3)Based on the simulation platform of containernet,this thesis verifies the advantages of GNT-MCTS-VNE and NIE in terms of delay and reliability over known methods GRC,MCTS-VNE,etc.