Dynamic sharing of resources between different Kubernetes clusters

As the adoption of Kubernetes continues to grow, so does the need for efficient resource utilization across diverse clusters, both within and across organizational boundaries. The Liqo framework introduces a novel paradigm by enabling Kubernetes multi-tenancy in a liquid, peer-to-peer model, presenting a promising solution for sharing computing resources across institutions. This thesis explores the concept of dynamic resource sharing between different Kubernetes clusters, laying the theoretical and practical foundation for broker platforms in the Liqo peering model. This thesis was developed in conjunction with the Liqo core team at Politecnico di Torino and stakeholders at TOP-IX – Torino Piemonte Internet Exchange. Seeking to further the common goal of efficient and seamless resource exchange, the result of this work was the design, implementation, and evaluation of three novel resource broker architectures, responding to different use cases and federation scenarios. Our proof-of-concept implementations exhibits promising performance and latency characteristics, showing that Liqo is a mature multi-tenancy solution that can adapt well to more centralized, large-scale setups. In addition to the development of resource brokers within the Liqo framework, this thesis extends its focus to the broader landscape of SaaS brokers. We review solutions in the GAIA-X and data spaces ecosystems, giving an insight into the evolving landscape of data sovereignty in modern, federated cloud-native environments. The findings and solutions presented in this thesis not only serve as valuable contributions to the Liqo framework but also provide a bridge between the concepts of Kubernetes cluster federation and the emerging paradigms of data spaces and GAIA-X. Overall, this work aligns with the broader industry trends of fostering data-driven innovation, collaboration and trust, realizing the full potential of cluster federation in modern cloud-native ecosystems.