Research On Offloading And Resource Allocation Technology Of Mobile Edge Computing

The number of mobile Internet users has grown rapidly.The needs of users are getting harder to meet.For this reason,researchers have proposed mobile edge computing technology.However,most of the user groups targeted by existing network resource optimization solutions belong to the same priority.Therefore,this paper divides the priority of user groups.In order to ensure the calculation of the main user tasks,while ensuring the use experience of secondary users as much as possible.This thesis designs corresponding offloading and resource allocation strategies for dual-priority scenarios.First of all,a parallel neural network algorithm based on computing interference ratio and content popularity is designed.This thesis uses the CPU frequency allocated to each priority task on the MEC server to set up the parameter to calculate the interference ratio.This behavior enables MEC to take into account the service experience of unauthorized users while ensuring the execution of tasks for authorized users.In addition,this paper discusses the transmission,calculation and cache model of the single-node dual-priority MEC system based on the user's system revenue during transmission,calculation and caching.Next,this thesis models the system model as a nonlinear optimization problem with integer programming.And a parallel neural network algorithm based on calculating interference ratio and content popularity is designed to solve the problem,which improves the profit of the MEC system.Secondly,a semi-supervised learning algorithm is designed.This thesis establishes a task queue,making the user's task queue a queuing model.In this way,the MEC server prioritizes the tasks of authorized users to ensure that the tasks of authorized users are not interfered by the tasks of unauthorized users.In addition,this thesis discusses the preemptive and non-preemptive models of the MEC server's handling of unauthorized user tasks.And this thesis also discusses the transmission and calculation model of the single-node MEC system based on the user's system revenue during transmission and calculation.After that,this thesis models the model as a nonlinear optimization problem with integer programming,and designs a semi-supervised learning algorithm to solve the problem by referring to the ? greedy algorithm.Compared with the previous algorithm,the algorithm solution delay is reduced,and the benefit of the entire system can be maximized.Thirdly,the worst response time resource allocation algorithm based on game theory is designed.This thesis sets reserved resources on the edge server to ensure that unauthorized users can use them without disturbing the uninstallation of authorized users.In addition,this paper starts from the time delay cost of user tasks,and constructs a game theory model with priority to solve the unloading decision.And this article also uses the worst response time of the task to allocate computing resources for the task.This algorithm overcomes the dependence of MEC nodes on user information when using a centralized algorithm for offloading decision-making and resource allocation.In summary,this thesis designs the above three algorithms for offloading decisionmaking and resource allocation of mobile edge computing.This thesis implements the service module of the algorithm,and simulates and analyzes the algorithm respectively.Experiments have proved that the average execution delay of the first algorithm is relatively large.The average execution delay of the second algorithm is significantly reduced.The execution delay of the third algorithm is greatly reduced,but the success rate of user task offloading is sacrificed.