Research On Micro-service Deployment And Task Scheduling Strategy Based On Collaboration In Edge Network

With the development of mobile communication technology and artificial intelligence technology,intelligent services such as augmented reality,the industrial internet of things,and the metaverse emerge steadily.The development of these services have caused the explosive growth of applications and mobile data,which put forward huge computing,storage and communication requirements for mobile communication networks.The currently deployed network is no longer able to meet these resource requirements.To meet the above challenges,edge cloud networks have been proposed and have received widespread attention.The edge cloud network combines the advantages of cloud computing and edge computing.It becomes an effective means to meet the needs of new intelligent services by uniformly scheduling the computing power,storage,communication and other resources through the collaboration strategy.However,considering the heterogeneity and resource limitations of edge nodes,the complexity and variability of network environments,and the diversity of users’ needs make it more complex to set collaboration strategies in edge cloud networks.A well-designed collaboration strategy can efficiently utilize heterogeneous resources in edge cloud networks,provide users with higher-quality services,and achieve higher system efficiency.Therefore,it is very important to design feasible and effective collaboration strategies based on specific network scenarios and task characteristics.This thesis mainly studies collaborative micro-service deployment and task scheduling schemes to optimize user latency,system energy consumption and the load balance of the cluster resource in edge cloud networks.And this thesis proposes the edge-edge collaboration based micro-service deployment strategy and the energy-aware based online task scheduling strategy.First,an edge-edge collaboration based micro-service deployment strategy is proposed to minimize user latency.Firstly,cloud-native technology is introduced into edge cloud networks.To minimize user latency and balance cluster resource load,a micro-service deployment optimization problem is established.This problem considers the impact of micro-service correlation and user distribution on service latency.Secondly,a distributed deep reinforcement learning deployment algorithm is designed to solve this problem,which enables each node in the cluster to make micro-service deployment decisions.Finally,simulation results show that the proposed deployment strategy can optimize user latency while ensuring resource utilization balance.Second,an energy-aware based online task scheduling strategy is proposed to minimize system energy consumption.Firstly,in the scenario of edge-device collaboration considering backhaul links,a system energy consumption optimization problem is established through task scheduling,which jointly considers task delay constraints and resource allocation of edge nodes.Secondly,an online asynchronous advantage actor-critic based task scheduling algorithm is designed,which can make online real-time scheduling decisions through multi-agent asynchronous learning.Finally,simulation results show that the proposed scheduling strategy can optimize the long-term average energy consumption of the system while meeting the delay constraint of the tasks.