Exploring Cost-Efficient and Resource-Optimized Kubernetes Deployments in Multi-Region Settings

Kubernetes is the predominant Cloud container orchestration platform. However, the default Kubernetes scheduler is cost-blind and region-agnostic. This can give rise to unnecessary financial costs when running clusters in public/commercial clouds. Existing research primarily focuses on cost optimization for single-region, homogeneous Clouds. Supporting heterogeneous clusters with cost and resource utilization under hybrid (on-demand and pre-emptible) pricing models remains under-explored. In this paper, we propose an end-to-end, cost-efficient, and resource-optimized scheduling framework designed specifically for distributed Cloud environments. Experimental evaluations on the Google Cloud Platform demonstrate that our approach substantially reduces costs (45% below baseline on-demand pricing and 62% below baseline hybrid pricing), whilst improving resource utilization.