Research On Task Offloading And Caching Strategies In Mobile Edge Computing

Mobile edge computing(MEC)is the most popular computing paradigm in recent years.The popularity of 5G networks,large-scale applications of the Internet of Things,and massive cloud service scenarios have kept MEC research hot.The development of technologies such as augmented reality,unmanned driving and the integration of spaceair-ground network means higher requirements for MEC-related research.Compared with the traditional Mobile Cloud Computing(MCC),MEC has a distributed architecture and the server is close to the end user,which can meet the user’s requirements for time delay.At the same time,MEC is expected to solve the response time requirements,limited battery life,excessive bandwidth costs and other problems.This thesis first studies the joint offloading and resource allocation algorithm in the MEC offloading and caching system with single MEC and multiple users.In this system,there is only one MEC,and this MEC is actually the MEC server of the base station,also known as the core MEC.Mobile wireless users are distributed within the communication coverage of the base station.The tasks of wireless users can be calculated locally,unloaded to the MEC server,or coordinated by other user devices.This thesis studies the communication model of multi-user interference in this scenario,minimizes the energy consumption of all mobile wireless devices by jointly optimizing the unloading decision and resource allocation,and designs a two-level optimization algorithm to solve this problem.The simulation results show the advantages of the two-level optimization algorithm in energy saving.Then,this thesis studies the service placement algorithm in the MEC offloading and caching system of multi-MEC and multi-user.In order to better assist the system in service placement on the MEC server,this thesis introduces a digital twin network to help the cloud and MEC server better respond to the service requests of wireless device users.Network operators need to minimize the load on the cloud or maximize the service requests of users from MEC servers.Based on this optimization goal,this thesis designs an improved genetic algorithm based on Merkle tree to place services.Simulation results verify the effectiveness of the algorithm.Finally,this thesis designs a simple physical simulation platform for the single MEC multi-user MEC offloading and caching system and the multi-MEC multi-user MEC offloading and caching system.Use the open source network simulator ns3 as the network framework,and raspberry pie as the deployed hardware carrier to build the physical platform.The effectivenesses of the proposed two-level optimization algorithm for single-MEC multi-user MEC offloading and caching system,and the service placement algorithm for multi-MEC multi-user MEC offloading and caching system are verified on this platform.