Research On Resource Scheduling Load Balancing Based On Kubernetes Cluster

The emergence of container orchestration tools is critical to the application of containerization technology in cloud computing,which greatly reduces the diffi culty of resource management and scheduling distribution in cloud platforms.The scheduling algorithm of the container orchestration tool is the key to container cloud real-time response to resource requests and improve resource utilization.Therefore,designing effective resource scheduling algorithms has become a key research topic in the field of container orchestration.The thesis proposes the corresponding optimization strategy for the deficiency of the mainstream container orchestration tool Kubernetes scheduling algorithm.The main work is as follows:(1)In the establishment of the Kubernetes basic scheduling model,considering the impact of the resource utilization of a single node on the overall load of the cluster,the remaining resource utilization of the node and the load balance degree are weighted as the node optimization algorithm.Insufficient resources in the cluster can easily cause Pod scheduling failure.This paper establishes a priority scheduling queue model based on the pod scheduling time and the amount of resources required,which effectively reduces the problem of repeated failures in pod scheduling.(2)This paper establishes a Kubernetes scheduling model based on cuckoo search algorithm.The basic scheduling model only considers the best performance of a single node in the optimization stage.The paper combines the scheduling time of the Pod sequence and the overall load balance of the cluster to obtain a multi-objective optimization function.Based on the objective function,a cuckoo search algorithm model is established,and the existing heuristic algorithm model is introduced for comparative analysis.The experimental results show that the search results of cuckoo search algorithm are better than the existing methods.The cuckoo search model can be used for Pod scheduling,which significantly improves the load balance of the cluster.(3)In order to optimize the optimization time of cuckoo search scheduling model,this paper proposes an improved cuckoo search algorithm.Chaos mutation is introduced to update the better solution through the standard node optimization process to obtain the initial population with better dispersion and continuity.Introduce reverse learning in the population iteration to expand the search space and increase the diversity of solutions.By analyzing the results of simulation experiments,the improved cuckoo search algorithm has improved convergence accuracy and has fewer iterations.Utilizing Go's native support for coroutines,the cuckoo bird population was divided into several sub-populations,and the sub-populations were optimized on multiple Goroutines in parallel.Finally,Pod scheduling was performed on the master according to the optimal solution(4)The resource scheduling results of different models in Kubernetes cluster are compared,and the experimental results show that the scheduling model established by the paper is better.In addition,a pressure tester was introduced into the cluster for service performance testing.The test results show that the improved Cuckoo search scheduling model enables the Kubernetes cluster to have better service capacity and shorter response time.Finally,the paper summarizes the research work of Kubernetes resource scheduling algorithm,and looks forward to the follow-up work.