Research On Edge Server Placement Strategy For User Plane Function

As 5G communication technology becomes more widespread,the global data volume ushered in a new round of explosive growth.However,cloud computing,the current mainstream computing paradigm,has been unable to carry and handle such a sharply increasing data traffic.For this reason,edge computing emerged as a new solution.Edge computing provides users with close-distance computing and storage services by deploying servers at the edge of the network,thereby reducing network latency and network bandwidth pressure.Among them,UPF(User Plane Function)is the link between the 5G core network and the edge network,which makes it important to consider UPF in edge computing.Most of the existing research on edge computing focuses on the traditional threelayer edge network model,and optimizes the edge computing solution on this basis,while ignoring the impact of UPF devices on the edge network in the 5G network model.Therefore,this thesis focuses on the research on the joint placement strategy of edge servers and UPFs in 5G scenarios,and is committed to providing users with lowlatency,high-availability services with limited costs,and on this basis to balance the load among various devices difference and prolong its service life.The main research content of this thesis includes:(1)This thesis proposes a genetic algorithm(CAGA)based on improved Canopy clustering for UPF and edge server placement in static scenarios.The algorithm first uses the load-based improved Canopy clustering algorithm to roughly cluster the data points of the base station to determine the number of edge servers;secondly,it adopts a fixed number of UPF placement strategies that are co-located with the edge servers,and the user access delay and Edge server load balancing is modeled;then the multiobjective optimization problem is transformed into a single-objective optimization problem using weight coefficients;finally,the genetic algorithm is used to find the optimal solution for this single-objective optimization problem,and the optimal placement strategy is obtained.Using real datasets for comparative experiments,the results show that the proposed algorithm reduces access delay by an average of 4.3ms compared with K-means,and reduces the load balancing by an average of 14%compared with Top-k.(2)This thesis proposes a dynamic UPF and edge server placement algorithm(KMNSGA)based on NSGA-II for the UPF and edge server placement problem in dynamic scenarios.The algorithm first uses the K-means algorithm to cluster and initialize the data points of the base station;secondly,uses the greedy UPF placement strategy to obtain the current optimal UPF placement scheme;then uses NSGA-II to analyze user access delay,edge server load Optimize load balancing and UPF placement costs and obtain a set of better placement solutions;finally use TOPSIS to filter out the optimal placement solution.Using real data sets for comparative experiments,the experimental results show that this algorithm can reduce 3.6% user access delay,9% edge server load difference and 5% UPF compared with K-medoids while ensuring UPF load is relatively balanced Placement costs.