There are a number of virtualization techniques that 5G can use. While it’s not quite ready for prime time, a cloud-native architecture is a front runner.
Telecommunications companies have been leveraging virtual-machine-based (VM) cloud platforms for the last few years. But as the industry ramps up 5G network installations, there are latency and operational efficiency challenges that need to be considered.
5G requires much higher throughput than previous generations (i.e., 20 Gbps for 5G compared to 1Gbps for 4G) and substantially lower latency (i.e., 1 ms compared to around 50 ms for 4G). Moreover, the massive machine-type communication of 5G requires an agile network with high availability.
This is where the cloud-native architecture (CNA) comes into the picture. Cloud-native is a microservice-based architecture where each service is containerized and managed independently. It solves many issues for service providers to deploy next-generation 5G commercially. Yet CNA functions cannot address some important challenges including networking and network uptime, which makes it hard to lock them in as the best choice for 5G rollouts.
What’s the solution?
5G is a hot topic everywhere these days. With a wide range of services, 5G rollout is a priority for both service providers (SPs) and network equipment manufacturers (NEMs) around the world. They are all seeking the most suitable underlying environment to host this heterogeneous technology and launch a standardized platform into the market as soon as possible. Some major SPs such as AT&T, Verizon, and Vodafone have already deployed commercial 5G networks with limited functionality. For example, AT&T and Verizon have installed 5G service in dozens of states in the US, and Vodafone has installed 5G in five European countries.
Historically, cellular network functions were deployed as independent boxes that were managed and manufactured by NEMs. Then around 2013, they were transformed from monolithic networks to SDN/NFV-based virtual networks that were more agile and delivered new revenue streams and lower operating costs for service providers. While SDN/NFV technologies are still being used, they are not leveraged as much as needed for 5G revolution. Still, there are certain issues that persist in terms of the current ways of using them. For example:
- The “noisy-neighbor” issue—where one application consumes a large volume of resources leaving very little left for other applications hosted on the same server—impacts network performance at large.
- The monolithic approach of hosting applications is an inefficient use of resources.
Considering its use cases, 5G needs a more enhanced cloud model than the one used for 4G and previous generations. With 5G, subscribers benefit from high-speed internet with ultra-low latency and new radio (gNB). This will open up a whole new world of innovative services, including enhanced augmented reality, smart-city services, connected vehicles, enhanced wearables, and much more.
To deliver these services, SPs need a rigid, scalable network capable of hosting an SDN-based network-slicing solution. The key objective of network slicing is the service-based architecture (SBA) of 5G, which demands policy-driven service selections for subscribers. SBA often requires scaling of a particular component of network functions in real-time. However, scaling the whole application to satisfy the needs of its subcomponents wastes resources and impacts network performance as part of the application remains unused when scaling.
How can SPs solve this scaling puzzle? With a cloud-native architecture.
CNA allows deployment of network functions as a cluster of containerized microservices, where each microservice can be deployed, scaled, and upgraded independently. Instead of scaling the whole application, only the required component within the VNF is scaled. There are many cloud-native orchestrators in the market—for example, Docker Swarm and Kubernetes—that can meet the rapid-deployment objective of 5G. And there are other advantages of using CNA for 5G implementations in the telco cloud:
- Prevent vendor “lock-ins” – Container-as-a-Service (CaaS) is a CNA-based service model that operators can employ to prevent vendor dependency.
- Self-healing – SPs can leverage the self-healing feature of CaaS for container recovery.
- Rapid and efficient deployments – CaaS implemented with DevOps pipelines can ensure continuous delivery and deployment of the services in production with automated build, validation, and deployment.
- Lightweight – The nature of containers can save a lot of resources which makes CaaS deployment and recovery faster.
CNA proves to be the best virtualization technique to drive telco innovation, especially for 5G. However, there are certain limitations and gaps that still need to be addressed before SPs can leverage this solution. These include:
- CNA was preliminarily designed for enterprise clouds where developers write and test applications. However, these environments are subject to long interruptions as a result of applications that are designed with downtime in the running environment. Therefore, CNA is not yet ready for tackling extremely low downtime—only about 5 or 6 minutes a year—of the telco cloud. Communities like the Cloud Native Computing Foundation (CNCF) and Linux are actively working to overcome this challenge of CNA’s vulnerability in continuous uptime.
- Kubernetes, the most widely adopted orchestration tool, still lacks some fast datapath networking capabilities such as ECMP, GTP tunneling, SCTP, and LACP. CNCF has started a project called network service mesh to support such features by Kubernetes. However, the project is still in development.
- Sharing operating systems could make containers or applications interfere with each other and can lead to resource contention, which could result in comprised latency, security, and efficiency.
In summary, cloud-native architecture can be a better alternative than other virtualization techniques for 5G implementations. But there are a few gaps and limitations that need to be addressed before it becomes fully compatible with the 5G architecture. With continuous efforts fromCNCF, Linux, and other communities, we can surely expect to have a CNA model that will be 5G compatible.
Connect with an Altran expert for more information about the advantages of a cloud-native architecture for 5G implementations.