Research On The Task Offloading Strategy Based On Edge Computing In The Tnternet Of Things

In recent years,the popularization of the Internet of things and the development of new computation-intensive mobile applications such as augmented reality and smart city have led to a large amount of data generated by terminal devices.Sending these data directly to cloud data centers would place a heavy burden on the current backbone network.In addition,the distance between the terminal and the data center can cause long transmission delays,resulting in poor user experience.By deploying computing and storage resources on the edge of the network,mobile edge computing can meet users' low latency requirements and relieve the pressure on the core network bandwidth.At present,mobile edge computing has been widely studied in caching,but it still faces some challenges in offloading decision and resource management.For example,the limitation of computing and storage resources of edge servers is ignored in offloading strategies,which affects the system performance.A large amount of task needs a long queue delay.In addition,in the multi-edge node cooperation strategy,there is a lack of distributed optimization scheme,the scope of cooperation between nodes is small,and the potential of collaborative computing is not fully exploited.Based on the above challenges,this paper starts from the two aspects of task offloading of a single edge server and cooperation of multi-edge nodes based on the related research.The main work includes the following:(1)In view of the limited computing and storage resources of a single edge server,a large amount of tasks can cause long queue delay,which will affect the user experience,a three-layer computing framework is constructed.In this framework,the terminal device will upload the sensing to the edge server,which will choose the appropriate ratio to offload,and upload the rest to the cloud data center.Firstly,a mathematical model is established with the goal of delay minimization under this framework.Under the constraint of energy consumption,the optimal off-loading ratio of edge servers is solved by using the sequential quadratic programming algorithm.Secondly,through the MATLAB simulation experiment,the influence of different arrival rates on the optimal unloading ratio and access delay is analyzed,and the proposed solution is compared with the benchmark scheme.The experimental results show that the proposed scheme has the best delay performance when the load arrival rate is much higher than the processing rate of the edge server.(2)In order to solve the problem of high energy-consumption,large transmission delay and serious signal interference caused by the frequent interaction among the large-scale multi-edge nodes,an edge node cooperation scheme was proposed for the offloading ratio maximization.Firstly,it analyzes the relationship between two performance indicators under the three-tier network architecture: the success rate of task offloading and the collaboration cost.A mathematical model was established for the offloading success ratio,the optimal solution of the problem was obtained by using Lagrange multiplier method under the constraint of the collaboration cost threshold,and the transmission delay of the collaboration and non-collaboration was analyzed.Secondly,in order to carry out cooperative offloading without the private information of each edge node,a distributed variant is proposed to solve the optimal cooperative size based on the classic ADMM.After obtaining this value,a forwarding tree is constructed to organize cooperative edge nodes.Each edge node forwards the matched task offloading request to its partner along the branch of the tree with the minimum interaction cost,and the implementation steps of each edge node are discussed.Finally,through the experimental simulation,the influence of various parameters on the success rate of offloading was analyzed,and non-cooperative and cooperative schemes were analyzed.The experimental results show that the offloading success rate of collaboration scheme is improved by about 20% compared with that of non-collaboration.