Research On Dynamic Deployment Strategy Of Edge Services Based On Deep Reinforcement Learning

With the rapid developing of mobile communication technology,more and more devices are connected to the edge of network and so that more and more data generated.Therefore,people start to distribute the computation tasks to machines at the network edge and build a novel and complete mobile edge computing(MEC)architecture.Due to its outstanding properties of low latency,context-aware,security,privacy protecting and energy saving,it was regarded as the key technology of next generation.Service deployment problem is the basic problem in creating a edge service provision system.In this paper,we monitor the requests of mobile users in different time period and consider the resource capacity,serving area and locations.By modeling the process with Markov decision process,we design and propose a Deep Deterministic Policy Gradient(DDPG)based algorithm to deploy services dynamically on edge servers so that the performance of system can be improved.By conducting a series of experiments on the YouTube request dataset comparing to the baselines,we show that the performance can be improved 10%at least.