A Novel Framework for Cross-Cluster Scaling in Cloud-Native 5G NextGen Core

Abstract

Cloud-native technologies are widely considered the ideal candidates for the future of vertical application development due to their boost in flexibility, scalability, and especially cost efficiency. Since multi-site support is paramount for 5G, we employ a multi-cluster model that scales on demand, shifting the boundaries of both horizontal and vertical scaling for shared resources. Our approach is based on the liquid computing paradigm, which has the benefit of adapting to the changing environment. Despite being a decentralized deployment shared across data centers, the 5G mobile core can be managed as a single cluster entity running in a public cloud. We achieve this by following the cloud-native patterns for declarative configuration based on Kubernetes APIs and on-demand resource allocation. Moreover, in our setup, we analyze the offloading of both the Open5GS user and control plane functions under two different peering scenarios. A significant improvement in terms of latency and throughput is achieved for the in-band peering, considering the traffic between clusters is ensured by the Liqo control plane through a VPN tunnel. We also validate three end-to-end network slicing use cases, showcasing the full 5G core automation and leveraging the capabilities of Kubernetes multi-cluster deployments and inter-service monitoring through the applied service mesh solution.

Keywords

• multi-cluster
• liquid computing
• multi-cloud
• network slicing
• vertical and horizontal scaling