Research On Task Unloading And Resource Optimization For 5G Edge Computing

As 5G moves from concept to application,architectures such as distributed networking,edge-independent networking,and ultra-dense networking have begun to be piloted in the industry.Distributed networking provides industry hard slicing based on operators’ 5G networks for industrial customers(such as power,police,etc.);edge independent networking provides relatively closed 5G private networks for industrial and medical customers who need security and isolation;Dense networks are based on 5G’s relatively higher spectrum usage and a large number of small cells.The three network structures will introduce a large number of 5G data plane network elements(UPFs).The dense deployment of UPFs provides attachment points for MEC servers,where the network and computing power are closely integrated.While providing ultra-low latency and large bandwidth for massive terminals with network access,it also provides network-as-storage and network-as-computing power services.With the help of such a network structure,mobile terminals can offload computing tasks to the computing power of the edge side,which will be a huge challenge to the system architecture.How to offload the computing tasks of mobile terminals and resource allocation of edge computing servers under the 5G SA network will be one of the key issues for the large-scale commercial use of mobile edge computing.This thesis briefly examines the effectiveness and pertinence of the algorithm implementation and considers the operator’s complex network environment and the overall architecture of specific application services.Instead,it first analyzes and simulates the algorithm based on the network environment of an independent park.Ideal for experimental testing and verification in the MEC environment.In the 5G+MEC private network scenario of an independent campus,this thesis designs a strategy that takes into account both computing task allocation and MEC resource allocation.Firstly,a computing task offloading model is established,and the system cost is simplified as the weighted sum of delay and energy consumption,and the complex problem of computing task offloading and MEC resource allocation is transformed into the problem of minimum overall cost.And it is divided into two sub-problems solving: resource allocation optimization and computing task offloading.The resource allocation optimization problem is solved based on the Lagrangian multiplier method,and a decision-making scheme for computing task allocation is obtained based on the greedy algorithm optimization.The final solution is obtained through the bimodal optimization algorithm.The final simulation results also show that the designed algorithm significantly reduces the system execution cost(compared to other benchmark algorithms).In the construction of a 5G+MEC private network in a power plant,it includes a dedicated 5G core network,dedicated bearer Flexe,dedicated macro station and room branch construction,and dedicated MEC server construction.The overall network environment is an independent private network.At the same time,the 2C terminal has not entered the campus,and the overall verification network environment is relatively clean,which is close to the network environment of the decision-making scheme for computing task allocation designed in this thesis.It manages MEC resources through the operator’s MEP platform,conducts experiments on the switching of multi-channel live video cameras on the server side,and reflects the overall system through the delay.Finally,the scalability,limitations and future development directions of the model in this thesis are analyzed.