Research On The Pre-migration Algorithm Of MEC-based IoV Virtual Machine

In recent years,cloud computing has been popularized in many fields of computers.Cloud computing provides computing,storage and other functions through the network.With the development of the Internet,more and more demands are shifting to Mobile Cloud Computing(MCC),that is,cloud computing can be applied in a mobile environment.However,there are also some limitations of MCC,such as limited resources in mobile devices,network bandwidth and delay cannot meet the requirements,as well as security issues and so on.This requires the transformation of centralized cloud computing to mobile edge computing(MEC),and push mobile computing,network control and storage functions to the edge of the network.Owing to the closeness to the user terminal side,MEC has the characteristics of lower network delay and higher bandwidth,which is very suitable for a service environment that can directly access real-time network information.For example,the Internet of Vehicles scenario is a popular application scenario of MEC.However,there are huge challenges of the Internet of Vehicles based on MEC in mobility management,mainly reflected in: high-speed mobile vehicles means frequent application migration,resulting in high service delays that seriously affect the user experience;Virtual Machine(VM)Migration involves server selection,migration time optimization and so on,and the higher migration time of the current VM migration algorithm cannot meet current needs.Frequent service migration will cause a severe service delay.In order to ensure that the server can provide a stable service to the vehicle and improve the user experience,this thesis introduces the data mining algorithm into the mobile prediction scheme.According to the historical trajectory of vehicle movement to predict the possible future position,a new movement prediction scheme based on data mining algorithm is proposed.Based on the predicted location results and server switching technology,the server deployment is completed before the vehicle reaches the destination area,reducing the negative effects caused by the vehicle position change,thereby achieving seamless server switching.The simulation results show that the accuracy rate of the vehicle position prediction scheme proposed in this thesis is 89.88%,which realize theshort delay when switching the server,and ensure the continuity of vehicle application services.To solve the problem of VM migration,this thesis designs a VM selection algorithm based on Markov model,which is used to select overloaded servers and VMs to be migrated.Since network resources are changing in real time,in order to arrange VM real-time migration and prevent the longer migration time of consumption of a large amount of network resources during the migration process.This thesis proposes a VM migration algorithm based on reinforcement learning.By observing the current demand level of the network,it finds the best time to schedule VM migration during the network saturation or peak period,reducing the consumption of network resources.The simulation results show that the algorithm proposed in this thesis can achieve good results in the entire VM migration process.