一种基于负载均衡的Kubernetes调度改进算法

An Improved Kubernetes Priority Algorithm based on Load Balancing

Kubernetes是云计算领域中的容器技术编排工具和集群管理系统,默认加载的预选算法和优选算法能将Pod对象调度到集群中合适的节点中运行。但Kubernetes调度算法使用的资源模型仅包括了CPU和内存,未考虑节点的性能。此外,在优选过程中,对于未设置CPU或内存下限的容器,无论节点的性能如何,Kubernetes都采用相同的默认值。针对上述不足,基于负载均衡对Kubernetes调度算法进行了改进,实验结果表明改进算法能提高Kubernetes集群的均衡效率。

Kubernetes is an orchestration tool and a cluster management system based on container in cloud computing area. The default predicate algorithms and priority algorithms in Kubernetes scheduler can schedule pod to a suitable node to run. However, the resource model used by the default algorithms just contains CPU and memory, and doesn’t take the performance of node into account either. In addition, during the priority process, for the containers without the requests of CPU or memory, Kubernetes applies the same default values. Aiming at the shortcomings above, this paper proposes an improved priority algorithm based on load balancing. The experiment results show that the improved priority algorithm can enhance equilibrium efficiency.