Building a Serverless 5G Core Network on a Public Cloud

#Core Network

Yuji SEKIYA Laboratory (Graduate School of Information Science and Technology, The University of Tokyo) and SoftBank Research Institute of Advanced Technology have been conducting joint research on the Core Network (CN; the control unit of mobile communications) in the next-generation mobile system. In this joint research, we have been working on "Stateless Core Network" following a modern web application architecture adapted to cloud technology, to resolve the issues such as scalability and robustness that the existing CN faces. In this article, we present our stateless Core Network deployed on managed services of Amazon Web Services, Inc. (AWS).

Writer : Kunio Akashi, Yuji Sekiya (The University of Tokyo), Hiroki Watanabe, Katsuhiro Horiba (SoftBank Research Institute of Advanced Technology)

1. Different concepts of "Cloud Native" in the web industry and the telecommunication industry

For about the past ten years, the telecommunication industry has been advancing the virtualization of networks, known as Network Function Virtualization (NFV). Initially, NFV mainly involved Virtual Network Functions (VNFs) which provided software-based network functions in virtual machines. However, with the recent rise in popularity of container technology, Cloud-native Network Functions (CNFs) are becoming more common. It is said that with the introduction of Service Based Architecture in 5G Standalone (SA) technology, cloud-native network functions (NFs) are now realized, centered around microservices and container orchestrators, represented by Kubernetes.

It is important to note that the term "cloud-native" originally signifies "the ability to build and run scalable applications by fully utilizing the functionality of the cloud in system construction and operation", and does not necessarily mean the adoption or non-adoption of specific technologies [1]. In other words, just because technologies like containers, Kubernetes, or Service Based Architecture are adopted, does not necessarily mean that the entire application is cloud-native.

Looking back at the concept of cloud-native within the telecommunications industry, it's often interpreted as "building and operating Network Functions (NFs) using a collection of technologies referred to as cloud-native". In other words, there's a tendency to label Core Networks, constructed by NFs that employ technologies like containers and Kubernetes, as being cloud-native. However, this interpretation simply involves altering the technologies used to implement NFs - it does not necessarily guarantee the attainment of the scalability and robustness that CNs originally aimed for. Of course, by adopting containers and Kubernetes, certain advantages such as speed and flexibility in CN construction can be achieved, but this does not necessarily mean these networks exhibit all the characteristics originally associated with being "cloud-native". It could be argued that truly cloud-native systems – those that can fully harness the inherent scalability and robustness of the cloud – can only be achieved when the overall development, construction, and operating experience of a CN application is aligned to meet the requirements characteristically pursued by cloud-native approaches.

On the other hand, modern web applications implemented on today's public clouds are actively utilizing an approach called "serverless" to ensure not just higher scalability and robustness, but also cost-effectiveness and elasticity. Serverless is a computing model where cloud service providers manage the resources needed for an application, dynamically allocating resources required by the smallest unit of the application, the function. The billing model in serverless is not based on the amount of resources pre-allocated, but rather the amount of resources actually used. As a result, modern web applications that have adopted serverless can be said to have realized the original "cloud-native" system.

2. "Cloud-Native" Mobile Core Network following modern web services

We further evolved the architecture of the Procedure-based 5G Core (Proc5GC) that we previously proposed, and built a core network (CN) that truly realizes "cloud-native", using Platform as a Service (PaaS) features on the public cloud to implement CN. More specifically, we used AWS’s PaaS features such as the Amazon SQS for message queues, the Amazon DynamoDB for databases, and the AWS Lambda function for serverless architecture, to construct "procedures'' needed for processing requests from User Equipment (UE) by linking Lambdas, and perform responses.

In this architecture, the message queue absorbs requests from the client, and subsequent server-side processing is reactively called depending on the type of message - making the CN operate almost like a web application. There is a glue on Amazon EC2 for maintaining compatibility with standard 5G systems, but essentially, you can build a 5G-compatible CN simply by combining the AWS Lambda function program for processing signals and the managed services.

In this implementation, we did not combine various cloud services for awkward usages or introduce proprietary plugins not provided by managed services. The results of this research were submitted and accepted as a paper at the IEEE 38th International Conference on Information Networking (ICOIN 2024) [2]. We gave an oral presentation on this paper at ICOIN 2024, held in Vietnam from January 17th for three days in 2024. As a research that radically revises a CN architecture in cooperation with the telecommunication company and the university, it received a lot of attention and gave rise to an actively engaged Q&A session.

3. Future Prospects

We have named the CN architecture utilizing the public cloud as “Cloud5GC” and are conducting various validations as part of our research and development efforts. Cloud5GC is an architecture that allows us to fully leverage the advantages of scalability and distribution inherent to the public cloud, as well as the statelessness and fault tolerance of Proc5GC. It possesses flexible scalability, adapting to the size of the service, and robustness that remains unaffected by a single failure.

The Cloud5GC architecture was also presented as a keynote speech at the ETSI NFV#44 Telco Cloud Executive Roundtable, attracting interest as a new architecture among participants. Currently, we are preparing for experiments to operationalize Cloud5GC as the actual CN for 5G networks.


[1]. Cloud Native Computing Foundation (CNCF), “CNCF Cloud Native Definition v1.0”,

[2]. K. Akashi, S. Yamamoto, H. Sakurai, K. Ito, T. Ishihara, and T. Iimura (The University of Tokyo, Japan); H. Watanabe, K. Shima, and K. Horiba (SoftBank Corp., Japan); Yuji Sekiya (The University of Tokyo, Japan), “Cloud5GC: Design and Implementation of Scalable and Stateless Mobile Core System on Public Cloud”, In Procedure of IEEE International Conference on Information Networking 2024 (ICOIN 2024).

Research Areas