Designation And Implementation Of Reinforcement Learning Based Task Scheduler For Kubernetes Edge Computing Framework

Based on container technology and distributed computing technology,Kubernetes has been widely used in the field of edge computing by virtue of its lightweight,easy to use,elasticity and other characteristics.However,the edge computing task scheduling problem in Kubernetes has not been fully studied.In order to make full use of edge computing resources and organize computing tasks reasonably in Kubernetes,this thesis designs and implements a reinforcement learning task scheduler for Kubernetes edge computing framework based on reinforcement learning algorithm.This thesis first reviews the background of edge computing and the application of Kubernetes in edge computing,and introduces the current research on edge computing task scheduling.Then,this thesis analyzes the principle of Kubernetes and its task scheduling mechanism,and introduces the reinforcement learning algorithm,and the algorithm can play a role in the task scheduling mechanism of Kubernetes framework.In this thesis,the task scheduling problem of Kubernetes edge computing framework is investigated,and then the task scheduling problem of Kubernetes edge computing framework is modeled and transformed into Markov decision process.This thesis gives full consideration to the characteristics of Kubernetes edge computing framework,and uses the near end strategy optimization algorithm in reinforcement learning to design reinforcement learning task scheduling algorithm,including task number prediction sub-algorithm,automatic scaling sub-algorithm and container scheduling sub-algorithm.Then,this thesis uses the reinforcement learning task scheduling algorithm to design and implement a task scheduler for Kubernetes edge computing framework.Finally,in the simulation environment,the reinforcement learning task scheduler based on Kubernetes edge computing framework is tested.The results show that the performance of the task scheduler is better than that of the default task scheduler in Kubernetes,which can effectively reduce the total cost of task execution.