- Blog
- Network
Large-scale Verification of Mobile Core Utilizing Public Cloud
#Core Network
Jul 29, 2024
SoftBank Corp.
Blogs
In 2022, Dish Network, a major US telecommunications provider, made a groundbreaking move by building a commercial mobile network on AWS. This initiative marked the beginning of a new era for the utilization of public cloud in mobile networks. However, for existing telecom operators with tens of millions of devices, the question remains whether they can maintain their current service quality while leveraging public cloud solutions. In this article, we will explore the results of our large-scale evaluation, examining the performance of public cloud infrastructure in supporting a mobile core network that accommodates 10 million devices.
1. Advantages of utilizing public cloud in mobile networks and responsibilities of telecommunications companies
To date, various operational advantages of running mobile networks on public cloud platforms have been proposed. For example, optimizing costs through dynamic resource allocation using Pay-as-You-Go services, leveraging Infrastructure as Code (IaC) for advanced automation, and utilizing AI/ML models for network monitoring and optimization are among the many potential benefits.
However, no matter how many benefits there may be, the most important responsibility for telecommunications operators is the stable operation of mobile networks. For example, Japanese telecommunications companies that utilize public radio spectrums have the duty to promptly restore services and report to the relevant authorities in case of accidents that result in the suspension or deterioration of telecommunication service quality. (For example, in the case of voice transmission services handling emergency calls, if a service outage affecting more than 30,000 people for over an hour occurs, a prompt situation report and detailed report to the Ministry of Internal Affairs and Communications are required.)
SoftBank, while fulfilling its social responsibilities as a telecommunications operator, conducted an evaluation to assess the performance of a mobile core utilizing public cloud services, aiming to enjoy the numerous benefits that public cloud can bring to mobile networks. This evaluation, conducted with commercial mobile network operations in mind, involved deploying NEC's 5G mobile core on AWS.
2. Configuration of a mobile core utilizing public cloud
Figure 1 shows the overall configuration diagram. The NEC 5G mobile core used in this evaluation adopts an architecture known as “Cloud-Native”. In this architecture, the 5G mobile core is divided into three parts. The first part is the “Protocol Termination”, which connects to radio base stations called Radio Access Network(RAN) and telephone exchange switches called User Plane Function(UPF). The second part is the “Call Control Logic”, which processes actual call control signaling. The third part is the “Data Store”, which manages data (customer information and device status) handled by the “Call Control Logic”.
Such a “Cloud-Native” architecture may seem unusual to those familiar with the reference architecture standardized by 3GPP, as shown in Figure 2. In the 3GPP reference architecture, it consists of Network Functions (NFs) with various functionalities such as AMF and SMF. However, in a “Cloud-Native” 5G mobile core, these NFs do not operate on a single computer or software. The various functionalities that make up each NF are divided into fine-grained software called “Microservices”, which dynamically increase or decrease in number according to the required processing performance to adjust call control signal processing performance. In particular, the “Call Control Logic”, which corresponds to the Service Based Architecture (SBA) using HTTP from 5G, has a high affinity with the virtualized infrastructure technologies such as Kubernetes that support large-scale systems used in the cloud based platform. It is capable of auto-scaling based on the resource usage of microservices.
Furthermore, these 5G mobile cores can utilize managed services provided by public clouds, allowing them to apply a business model based on the “Pay as You Go” approach where resources are billed according to usage. For example, Kubernetes is used to run “Microservices” in the “Protocol termination” and “Call Control Logic” sections, while various managed database services provided by AWS are utilized in the “Data Store” section.
Figure 3 shows the configuration diagram of the verification conducted in AWS. Amazon Elastic Kubernetes Service (Amazon EKS) is used for container management with Kubernetes, and the “Protocol Termination” and “Call Control Logic” shown in Figure 1 are operated as Kubernetes Pods. Amazon EKS utilizes EC2 for the virtual machines of the Worker Nodes that run the containers. In the “Data Store”, Amazon Relational Database Service (Amazon RDS) is used for relational databases that store customer data, Amazon ElastiCache is used for high-speed in-memory data stores that store device states, and Amazon DocumentDB is used for Key Value Store that stores UE context and network slice information. These managed services provide services in a “Pay as You Go” manner, charging based on the resources used, including redundancy and operations. Therefore, from a 5G mobile core perspective, it eliminates the need for design and operation of non-functional requirements such as capacity planning and redundancy, allowing for a more focused approach on providing valuable services to customers and other important business matters.
3. Evaluation of a commercially-scaled mobile core utilizing public cloud
In this verification, assuming commercial operation in a mobile network, we evaluated the performance of a 5G mobile core utilizing the public cloud and confirmed its sufficient scalability and call control signal processing performance. Specifically, we built a 5G mobile core capable of accommodating 10 million devices and confirmed that it can achieve call control signal processing performance to register the location information of all devices and make them communicable within 30 minutes.
Furthermore, as part of the evaluation of the “Cloud-Native” 5G mobile core, we conducted an assessment of the auto-scaling feature. Specifically, we assumed situations with low demand for call control signal processing performance, such as during the night, and confirmed that the system can automatically scale out from a reduced state with 20% of peak resources to the point where it can provide peak performance in response to increased demand.
4. Outlook for the future
Through this verification, it has become clear that public cloud can provide the various advantages it offers even in the scale of commercial mobile networks. On the other hand, the current Telecommunications Business Act in Japan is currently under discussion for review and amendments regarding the full entrustment of telecommunications infrastructure, including to third parties such as public clouds, due to the development of virtualization technologies. Moving forward, we will actively participate in these discussions while continuing to challenge both the technological and regulatory aspects to ensure the reliable operation of mobile networks that customers can use with peace of mind.
Author: Katsuhiro Horiba