A Design of Policy-Based Scheduling for Federated Multi-Clusters

Kubernetes, the open-source container-based cloud orchestration platform, is becoming the de-facto standard for automatically deploying and managing containerized applications in the cloud environment. Recently, A cluster in Kubernetes can be extended to multiple clusters and the federated Kubernetes architecture is proposed for managing these multi-clusters. In this paper, we propose policy-based scheduling architecture for multiple clusters. The proposed scheduling architecture determines the appropriate federated clusters to deploy workload and the number of replicas to each cluster. The policy engine is used to determines the appropriate federated clusters to deploy workload, the number of replicas to each cluster, according to the given policies.