Microservice Mobility for 5G

Abstract

Microservice-based design has gained momentum also in 5G-and-beyond systems within the telco industry. For instance, in the 5G core, Service-Based Architecture (SBA) has been designed based on the microservice architecture for continuous integration and development using modern DevOps practices. The industry has been promoting Kubernetes as the de-facto platform for managing and orchestrating containerized network services smoothly and transparently. From a technical viewpoint, Kubernetes consists of individual worker nodes that can be used for running workloads. Multiple nodes can be organized into a cluster to increase the resources for running and scaling out workloads. However, Kubernetes has some technical limitations in the number of compute nodes that can be added to the same cluster. Furthermore, running all of the workloads of a large organization in a single cluster can pose a threat in case of failure. Fortunately, new extensions to Kubernetes currently support instances running in multi-cluster scenarios. In order to guarantee service continuity and service level agreements (SLA) in 5G-and-beyond, we have developed and benchmarked a prototype. The prototype implements seamless service migration across cluster boundaries for stateful microservices using Kubernetes Cluster Federation (KubeFed) and Network Service Mesh (NSM). It could be helpful to migrate services from one cluster to another, unlike container migration, which changes how the application works. We are implementing migration using a cloud-native way where the state of the microservice is transferred via Redis-based databases to a new cluster when the service needs to be moved to a new cluster. We believe such a mechanism can help migrate workloads from one edge cluster to another, such as providing services in close proximity to a moving mobile terminal for 5G and beyond.