Research On Edge Computing Network Task Offloading And Resource Allocation Algorithm

With the rapid development of Internet of Things technology,new applications such as autonomous driving,telemedicine,and smart construction sites continue to emerge.The traditional cloud computing architecture can no longer meet the requirements of these applications for low latency and low power consumption,then Mobile Edge Computing(MEC)has emerged as the times require.MEC technology deploys computing servers at the edge of the network to help users to calculate or store tasks nearby,which greatly reduces the waiting time of users and enables the network to support delay-sensitive tasks.How to reasonably allocate tasks to different MEC servers and effectively allocate the limited computing or storage resources of MEC servers to each task is also an urgent problem to be solved.Based on this,this thesis mainly studies the task offloading and resource allocation algorithm in MEC networks.Firstly,a delay minimization algorithm is proposed in MEC-enabled heterogeneous cellular networks with multi-users;then,for the multi-user cloud-edge server networks,taking into the buffer consideration of edge servers,a delay minimization algorithm and a cost minimization algorithm are designed respectively.The specific research content is as follows:1)For the MEC-enabled heterogeneous cellular networks which include multiple users,multiple micro base stations,and single macro base station,an optimization model is constructed with the goal of minimizing total delay of the system under the constraints of the battery capacity of user terminal equipment and the computing resources of MEC servers,and finally,a joint macro-base-stationmicro-base-station-terminal delay minimization algorithm is proposed.The algorithm decomposes the original optimization problem into two sub-problems with a sequence relationship: the micro base stations resource allocation strategy problem and the optimal task offloading decision problem.For the micro base stations resource allocation strategy problem,assuming that all the computing tasks on MEC server are known,the allocation strategy of computing resources assigned to each task can be solved by computing the KKT condition.For the offloading decision problem,by taking the micro base stations resource allocation strategy into the original problem,it can be transformed into an integer programming problem,and then the feasible solution for task offloading,which is also the sub-optimal solution of the original problem,is obtained through the branch-and-bound breadth-first search algorithm.Simulation results show that the proposed algorithm has better latency performance than other offloading schemes in different task scenarios.2)For the cloud-edge networks with multiple users,taking into the buffer of edge servers’ consideration,an optimization model is constructed with the goal of minimizing the average response time of tasks under the constraints of the stability condition of the queuing system,and finally,a probability-based delay minimization algorithm is proposed.By analyzing the feasible set and convexity-concavity identification of the optimization model,the optimization problem is proved convex,and the one-dimensional search algorithm called golden section method is used to solve the optimization problem.Compared with other task offloading schemes,the proposed algorithm has obvious advantages in reducing the average task response time.3)On the basis of 2),for the cloud-edge networks with multiple users,considering both the servers’ computing frequency and the capacity of buffers,an optimization model is constructed with the goal of minimizing the average response time and the total power consumption.Since the optimization problem introduces servers’ computing frequency and the goal of power consumption minimization,it is non-convex and cannot be solved by using the same algorithm in 2).Based on this,this thesis proposes a joint offload-probability and server-computing-frequency cost minimization algorithm.Firstly,an initial solution of the original problem is obtained by using the adaptive genetic algorithm.Then,the penalty function method is introduced to transform the original optimization problem to an unconstrained problem.And finally,the initial solution is substituted into the simulated annealing algorithm to obtain the feasible solution of the unconstrained optimization problem,which is a suboptimal solution of the original problem.This algorithm solves not only the problem of insufficient searching ability of genetic algorithm,but also the problem of dependence on the initial value of simulated annealing algorithm.The simulation results show that the proposed algorithm can converge fast and obtain a smaller system cost than other algorithms.