On Key Technologies And Applications Of Virtual Service Migrations Based On Reinforcement Learning

As a new generation of distributed computing infrastructure,cloud computing platform due to its photogenic for traditional platform in terms of performance and price advantage has become a hotspot in the research of the academia and industry in recent years.The sensitivity of access latency and the variability of the space during access is the typical feature of mobile cloud computing.From the user's perspective,it is important to maximize the availability of services.But as the number of user groups grows up and the network traffic increases,it becomes more difficult than ever before for cloud service providers.As such,providing services without considering these factors may significantly increase access latency,and much worse,increasing network traffic can also lead to outages and performance degradation.To mitigate this problem,it is an effective way to migrate the services to some of the user's advantageous locations in the network.Fortunately,with virtualization technology in the cloud,we encapsulate services in a set of virtual machines and migrate to one or different data centers as needed.Moving services closer to the user not only reduces service access latency,but also minimize the network costs for service providers.Therefore,this issue is very important for real-time services.However,service migration adds to the overall cost of services at the expense of bulk data transfers and possible service outages.To gain the benefits of service migration while minimizing service costs,in this paper,we leverage reinforcement learning(RL)methods to propose an efficient algorithm,called Mig-RL,for the service migration in a cloud environment.The Mig-RL utilizes an agent to learn the optimal policy that determines service migration status by using a typical RL algorithm,called Q-learning.Specifically,the agent learns from the historical access information to decide when and to where the service should be migrated,regardless of the service accesses.Therefore,the agent can dynamically adapt to the environment and achieve online migration in real time.The experimental results on the real and synthesized access sequences form cloud networks show that Mig-RL can minimize the service costs,and in the meantime,compared with existing algorithms,improve the quality of service(QoS)by adapting to the changes of mobile access patterns.