It means you want to add worker nodes to the OpenShift cluster on your on-premise environment. Is it right ? Then I don't recommend you the topology design for spreading your unusual loads. Because it's very difficult to keep very lowest latency which is required local network latency(<10ms) between two sites. It's biggest blocker to apply the design. Even if you can clear the latency, you cannot use the cloud provider features(dynamic storage provisioning and so on) you provided on public cloud. Additionally all pods should communicate with Router pods on the on-premise for external systems, it's tricky to manage the access path. So I suggest you to use multiple clustr management soultion instead of this design. Such as Red Hat Advanced Cluster Management for Kubernetes.