Design And Implementation Of Task Scheduling Management System For Edge Cloud Collaboration Platform Based On Kubernetes

In the face of the connection of massive user devices,the explosive growth of data traffic and the increasing demand of users for service quality,more and more enterprises begin to pay attention to edge computing.The rise of edge computing model does not mean the decline of cloud computing model.Edge computing can actually be regarded as a supplement to centralized cloud computing.The combination of the two can reduce latency,improve scalability,increase access to information,and provide users with better service quality.The process of deciding which node in the edge cloud cluster to execute computing tasks is called task scheduling.The quality of scheduling decision will directly affect the performance of the whole edge cloud cluster.Therefore,the design of scheduling algorithm is very important to the performance of edge cloud cluster.This thesis first studies the scheduling strategy adopted in Kubernetes,which is the current mainstream cluster management application.Analyzes the key factors that need to be considered in the task scheduling problem.Then studies the existing scheduling algorithms.Analyzes the advantages and disadvantages of various scheduling algorithms.Based on the above study,the demand analysis,architecture design and database design of the task scheduling system of the edge cloud collaboration platform based on Kubernetes are carried out.The functional modules are designed in detail,including communication module,policy configuration module,scheduling module,cluster monitoring module,load balancing module and result display module.Among them,scheduling algorithm is the core of the system.This thesis proposes an edge cloud collaborative scheduling algorithm for multi tasks based on ant colony algorithm,which comprehensively considers the edge cloud collaborative cluster resources,the network among cluster computing nodes,and the load ratio of edge cloud nodes.The scoring results are used to initialize the pheromone matrix,to avoid invalid search in the early stage of iteration.By introducing the dynamic random factors,the algorithm can better tradeoff the randomness of the algorithm in the early stage and the convergence speed in the later stage.Besides,the load ratio between edge and cloud nodes is added to the calculation of the objective function to ensure the edge cloud load ratio of the scheduling result.After each scheduling,load balancing prediction is carried out based on the scheduling results,which further reduces the decision delay of load balancing.Finally,the task scheduling system is deployed to the edge cloud collaboration platform.Through function test and performance test,the functional correctness of the task scheduling system and the effectiveness of various performance indicators are verified,which shows that the task scheduling management system of the edge cloud collaboration platform proposed in this thesis has good scheduling performance and operation stability.