Research On Optimization Strategy Of Resource Scheduler

Kubernetes is currently the most widely recognized Docker container orchestration system,and most cloud service providers use it as a deployment solution for cloud native applications.Although it has a very rich system function,as the orchestration system,the core of it is still the scheduling function.In this paper,the system component of the scheduling function in Kubernetes--the resource scheduler is taken as the research object,and the corresponding optimization strategy is proposed for the static and dynamic scheduling problems existing in the existing scheduler.First,the related technologies,such as Docker container technology,Kubernetes technology and cloud computing cluster scheduler technology,are studied separately.On this basis,the Kubernetes resource scheduler and its scheduling strategy are deeply analyzed,and the problems existing in the existing resource scheduler are discussed.Then,based on the re-optimization of the static scheduling process,a set of RRC(Reduce Redundancy Calculations)scheduling algorithm is proposed for the redundant computing problem of resource scheduler in static scheduling.This algorithm can reduce the redundant calculation steps when scheduling different copies of the same resource controller,thereby improving the scheduling efficiency of the scheduler.Then,for the problem that the cluster load may be unbalanced after static scheduling,a dynamic load balancing mechanism is proposed by analyzing its dynamic scheduling requirements.According to certain trigger conditions and changes in node load information,the mechanism can dynamically select an appropriate Pod for rescheduling to achieve load balancing.Finally,the Kubernetes cluster experimental environment is built and the experimental scheme is designed for verification.On this basis,data analysis is performed on the experimental results to verify the availability and effectiveness of the optimization scheme.The results of static scheduling experiments show that after the static scheduling optimization scheme is applied,the scheduler reduces the average scheduling time by about 9% when scheduling the Pod replicas for the same resource controller.The scheduler is optimized to show better scheduling performance;load balancing experiment results show that static scheduling can keep the cluster relatively balanced when the Pod is just scheduled,but the load will be unbalanced when the cluster is running for a long time.The dynamic load balancing mechanism proposed in this paper,combined with static and dynamic scheduling,can ensure that the cluster can maintain the balance of the cluster system well under the long-term operation,so as to maintain the stability of the cluster system better.