Auto-scaling, frequent deployment of services with a quick failure recovery (e.g., service availability > 99.999 %) are essential features to support 5G and beyond networks in cloud environments. Such service resiliency, flexibility and network automation in 5G networks as required for network slicing and orchestration, as well as services' evolution (e.g., service update/upgrade, scaling etc.), can be achieved via softwarization, virtualization, and cloud computing technologies. Aiming at delivering the applications as-as-service, cloud-native is a methodology of developing, building, and managing the applications that fully exploit the cloud computing model. Agility, self-healing and auto-pilot (e.g., auto-configuration), among others, are considered as the core features envisioned for 5G in cloud native environment.
There are three main design patterns for cloud-native approach. Firstly, the cloud-native applications are composed of microservices, where they can be composed of many services meshed together and operating independently of each other. Secondly, the cloud-native applications are packaged in one or multiple isolated containers, while managed by means of a set of standard APIs. Finally, they run in a continuous integration and delivery (CI/CD), where an application goes through fast cycles of development, build, test, release, and deployment.
Considered as one of the main features attracting the adoption 5G and beyond in cloud native environment, microservices and containers have small footprint and fast start times. There are many technologies for containerization, such as Linux Containers (LXC), containerd, CRI-O, docker, with a relatively similar performance. In order to cloudify the network, flexible functional split was introduced by 3GPP [TR 38.801] to cope with the ever increasing COPEX/CAPEX with the introduction of new services in 5G. Contrary to monolithic 4G RAN, 5G RAN is disaggregated into three main units: the Remote Radio Unit (RRU), the Distributed Unit (DU), and the Centralized Unit (CU). All the necessary components related to signal transmission/reception exist in RRU, while DU may contain a set of physical layer functions shifted to the cloud as well as some set of higher layer functions. The rest of higher layer functions will be then treated/processed at the CU. Similarly, the core network can be, monolithic (running as singly entity), disaggregated, or even re-aggregated for different objectives like optimize the resources, reduce the traffic on /access/mid/back-haul, etc. After shifting 5G functions to the cloud (i.e., build them as a cloud-native applications), the number of containers may grow drastically on per service basis (i.e., network slices and sub-slices). This calls for efficient management and orchestration of the ecosystem in order to achieve the envisioned performance. Several frameworks already exist, such as Kubernetes, Docker Swarm, and Apache Mesos. For the ease of packaging, deploying and managing Kubernetes applications, Operator framework is introduced under Kubernetes environment for encapsulating the lifecycle management operations, and thus facilitating service automation. Kue5G-Operator is a good example of such Kubernetes Operator, which will be exploited in this demo. It is worth noting that Kubernetes is considered as a defacto candidate for the orchestration of 5G and beyond in cloud environment as it has the largest community, better support, and thus long term support. The same also holds for Docker container model.
Motivated by the necessity of quick 4G/5G service deployed, re-configuration, and dynamic network management according to the ongoing user traffic, we present the current demo. In this demo, we present 5G dynamic network deployment, (dis/re-) aggregation, and (re)configuration in cloud-native environment. More specifically, we demonstrate in this demo the following: i) support 5G in cloud native via Kube5G-Operator and dynamically automate the network, i.e. upgrade/downgrade services, ii) change the configuration automatically, iii) dynamically switch between (dis/re-)aggregation modes for both CN and RAN according to user traffic. Note that the aforementioned features of Kube5G- Operator spans different phases of openshift operator, ranging from basic install (e.g., application provision), to auto-pilot (e.g., auto-configuration).