Research On The Pre-migration Strategy Of MEC-based IoV Applications

Mobile Edge Computing(MEC)provides an edge network environment with IT and cloud-computing capabilities for mobile terminals such as connected vehicles.It is one of the key technologies of 5G with low latency,local sensing and many other advantages.Internet of Vehicles(IoV)builds the new industrial value chain based on MEC.Through the application deployment localization,IoV can achieve the acquisition and transmission of travel data in a very short time,which may make great contributions to “smart transport”.The IoV applications based on MEC can bring the enormous changes to the existing network environment,and receive extensive attention from the field of the industry and academia.However,IoV applications based on MEC present a huge challenge in mobility management.While MEC brings great convenience for vehicles,it also proposes mobility requirement for applications deployed in edge networks,that is,the migration of IoV applications between different MEC servers.The scenario of Internet of Vehicles complicates the problem on mobility management.The high-speed of connected vehicles means more frequent application migration,resulting in large migration delay,which seriously affects the vehicle driver's experience.Besides,the IoV has developed and the relevant applications include safety,convinience,Internet and other services.Existing migration strategies of applications cannot meet their diverse performance requirements.Aiming at the problem that frequent migration causes a large number of migration delay,this thesis proposes a pre-migration method based on vehicle trajectory prediction.The core part of this method is to predict the future trajectory of the connected vehicles.Based on the predicted results and the virtual machine online migration technology,IoV applications can be deployed predictively on the target place before the connected vehicles reach,which reduces the downtime caused by the migration.The method adopts extreme learning machine as the algorithm of mobility trajectory prediction,which can get the prediction result very quickly.The simulation results show that the proposed trajectory prediction method can accurately predict the future direction of the connected vehicles with the accuracy of 93%,and greatly reduce the application downtime and service delay.In view of the diversified performance requirements of IoV applications,this thesis proposes a multi-application migration strategy,which dynamically allocates computing and network resources according to the requirements of applications.The dynamic programming algorithm is used to find the migration strategy firstly.At the same time,the negative effects caused by trajectory prediction are considered,and the evasive algorithm of negative effects is proposed.Then the priority of the IoV applications is defined by integrating various performance indicators.Finally,combined with the evasive algorithm,a multi-application migration algorithm based on priority queue is proposed.The simulation results show that the proposed algorithm is superior in terms of service stability,avoiding negative effects and meeting the performance requirements of various applications.