Design And Implementation Of Dynamic Service Placement And Service Migration Path Optimization Algorithm In MEC

In mobile edge computing(MEC),small coverage of edge nodes and limited number of deployable applications of edge node can lead to discontinuity service problem while terminal devices are moving.MEC supports relocation to ensure the seamless service.However,relocation process leads to additional delay and migration costs.To decrease delay,there are two moves:(1)predicting the user requests' pattern;(2)improving the availability of base station connection.To decrease migration costs,dynamic service placement strategy can be used.Therefore,how to improve the accuracy of prediction,load balancing performance of network and to decrease migration cost is becoming an important problem of relocation technology.To improve accuracy of prediction and load balancing performance,this paper proposes NRDPUC algorithm,which includes KALMAR mobility prediction algorithm and base station selection algorithm.KALMAR algorithm integrates Kalman Filter model and Markov model,which increases about 30%and 26.7%accuracy of the mobility prediction comparing with Kalman Filter model and Markov model.Base station selection algorithm first predicts the load of each base station and then classificates the base station according to the prediction result,this algorithm increases the load balancing performance comparing with NCRA algorithm.To deal with dynamic service placement issue,this paper proposes SMDQN algorithm.SMDQN improves DQN by optimizing action selection and Q-network update process.SMDQN algorithm determines the optimal location of service which can adapt to different user requests distribution and calculates an optimal migration path which can reduce the migration costs.The simulation result shows that SMDQN outperforms other related algorithms on migration costs,migration hops.This paper firstly introduces the state-of-the-art MEC researches,then introduces related technology in MEC and requirement analysis of dynamic service placement and migration path optimization algorithm;next introduces base station selection based on mobility prediction algorithm,dynamic service placement strategy and migration path optimization algorithm in detail.Simulation experiments and simulation results are shown respectively.Simulation results show that the performance of two algorithms is superior to other related methods.