A Cluster Autoscaler Based on Multiple Node Types in Kubernetes

Kubernetes, as a production-grade container orchestration system, has been used in varieties of fields on a large scale. However, as used more and more widely, the appropriate selection of auto scaling schemes in kubernetes also faces considerable challenges in their respective application fields. Based on multiple node types with image pre-loaded, this paper proposes an improved automatic scaling scheme that combines the advantages of different types of nodes in the scaling process. The results show that, compared with the default auto scaler, it can improve the performance of the system under rapid load pressure and reduce the instability within the running clusters.