Optimizing peer-to-peer multi-cluster communications for Liqo-based multi-cloud deployments

The growing complexity of cloud-native infrastructure, particularly in hybrid and multi-cloud contexts, has led to the widespread adoption of Kubernetes as the standard platform for orchestrating containerized workloads. While Kubernetes offers robust mechanisms for managing resources within a single cluster, it lacks native support for seamless multi-cluster deployments. To bridge this gap, the open-source Liqo framework enables dynamic federation of Kubernetes clusters, allowing workload offloading, resource sharing, and secure cross-cluster networking. Despite these capabilities, Liqo’s default networking model remains consumercentric: all traffic between provider clusters must be routed through the consumer, introducing unnecessary latency, limiting effective bandwidth between providers, and creating a potential single point of failure. This architecture limits the scalability and resilience of Liqo-based infrastructures in real-world scenarios. This thesis introduces an architectural enhancement to overcome this limitation. A new Kubernetes Custom Resource Definition (CRD), ForeignClusterConnection, is designed and implemented, enabling declarative and automated provisioning of direct tunnels between provider clusters. A dedicated operator orchestrates the lifecycle of these resources, interacting with Liqo’s internal APIs and CLI tools to discover cluster metadata, manage network endpoints, and propagate connection status. The implementation, based on the Kubebuilder SDK, follows established Kubernetes operator patterns, ensuring high modularity and seamless integration. Experimental evaluation under simulated WAN conditions demonstrates improvements in latency, throughput, and system resilience when direct inter-provider communication is established. The proposed solution aligns with Kubernetes’ declarative model and GitOps principles, paving the way for advanced features such as dynamic endpoint rewriting, policy-driven routing, topology awareness, and monitoring integrations. It enhances the efficiency, robustness, and observability of Liqo-powered multi-cluster deployments, contributing to the evolution of Kubernetes-based federated networking in heterogeneous environments.