By Paul Miller
The promise and complexities of virtual 5G networks
While 5G is becoming a well-known technology buzzword, not much is known about the major shift the industry is going through as 5G deployment heads towards prime time. Obviously, 5G’s promise to the service provider community is one of increased revenue through an explosion of use cases such as Industrial IoT, autonomous and vehicle-to-vehicle (V2V) applications, mobile edge computing (MEC), human machine interaction such as augmented reality, and much more. Combined with the higher bandwidth and massive endpoint count support that 5G introduces, a perfect storm of opportunity has been created. Hence the incredible worldwide investment starting around transforming the globe’s communications to 5G.
However, as always there is complexity in the details. 5G, and its tightly coupled virtual radio access network (vRAN) initiative, combined with the new mobile applications mentioned above, creates the need for a physically distributed (geo-distributed) virtualized infrastructure onto which these applications – both core network functions and advanced new applications – can be hosted. This represents the first time in the history of telecommunications that such a geo-distributed virtualization approach has been selected for deployment.
This is largely due to the emergence of applications that are fully software based, with exclusively IP based communications, that are dependent on ultra-low latency (URLLC) access from the edge to the core to enable the application to function properly. To understand the criticality of this, imagine introducing too much latency into a V2V communication being use to vehicular accidents in real time.
The benefits of the virtualization approach are many – ultra-fast time to market with near instantaneous deployment, upgrade and replacement of virtualized applications spread throughout the network. Management of the entire far edge to core network via a single virtualized cloud infrastructure. Transformation of a monolithic, proprietary vendor environment to an open ecosystem; one based on open source technologies and best-of-breed disaggregation thereby driving cost down to build and operate the network. Decoupling the software from the silicon it operates on. No other approach contains the promise so intrinsic to this transformative technology.
The complexities surface when the realities of the service provider environment are added. This new network architecture will have tens of thousands of deployed compute nodes, within a single operator’s infrastructure. Operating, deploying, managing and debugging this deployment is a challenge so complex that new tools and technologies must be brought to bear to ensure feasibility.
Enter the 5G Cloud Native Solution: Wind River Cloud Platform
Combined with the requirements mentioned above, another parallel technology transformation in the foundation of virtualization is well under way. This is, of course, the transition from virtual machines to containers, as a much more performant, scalable, and resource efficient solution for virtualization. The industry knows this as Kubernetes (container orchestration) and Docker (the container runtime), which have created the underpinning of the revolution.
As a result, there exists a need to take the best of breed virtualization technology, Kubernetes and Docker, and using that as a foundation create an end-to-end cloud solution that enables and embraces the challenges of a geo-distributed network architecture. This must be done meeting both operational and technical requirements.
To do so successfully for 5G requires the separation of the vBBU (virtual Base Band Unit) into two functions, known as the vDU (virtual Distributed Unit, close to the RU or Radio Unit) and the vCU (virtual Central Unit, closer to the core). Compute nodes hosting the vDU 5G functions must be geo-located at the far edge, close to the RU. The vCU aggregates the thousands of vDU units, whose traffic is then passed to the Core network functions.
Wind River Cloud Platform has been designed from the foundation up for this specific challenge. The ability to scale down to a single node at the edge, while supporting ultra-low latency and onboard accelerators for real time radio applications is combined with the industry’s best Single Pane of Glass (SPoG) solution for managing the entire cloud.
The importance of scalability and zero-touch automation
Additionally, the ability to scale to thousands of distributed sites is important, each with varying levels of compute node scale, and to dynamically be able to change the scale at any site while the applications run uninterrupted in a 6 9’s protective environment.
Wind River Cloud Platform is the industry’s only solution that enables scale down to a single physical node at the far edge vRAN location, providing an unmatched CAPEX enablement to the service provider’s business case.
Any sub-cloud region must be able to be managed while applications are active. This includes not only the ability to scale the cloud site via additional nodes for workload or storage while the system is fully operational, but the ability to orchestrate updates and upgrades on the running infrastructure – dynamically relocating workloads as the software update process runs in a zero-touch full autonomous mode. This includes of course the ability to roll back any update dynamically and automatically if desired. All of these abilities are intrinsic to Cloud Platform, demonstrating the operator focused feature set that is woven throughout the solution.
Finally, zero-touch extends to the deployment and activation of new sites, where advanced automation technologies are used to enable a fully hands off, zero human involvement to turn up remote controllers, activate them, and enroll new worker nodes. This includes the ability to take a known good, reference far edge cloud configuration and deploy it in an automated fashion as new cloud instances are created. This is critical for an architecture with tens of thousands of sites using a common design.
Latency plays an ever-increasing role in the ability for 5G deployments to function, scale, and cost reduce. As a rule of thumb, 5uS represents approximately 1 kilometer of fiber. When the cloud platform selected for 5G deployment has improved latency, this can result in greater allowed distances between the RU and vDU, which then increases the pooling of vDU resources.
This significantly reduces the number of required nodes in a deployment, positively affecting CAPEX across compute node hardware, vRAN software, SDN, and site related costs. These economic benefits are all sourced from a highly efficient, low latency cloud technology. The fundamental cost of deploying a 5G network is heavily influenced by the efficiency of the cloud infrastructure selected for it.
When applications are considered, such as those enabled by URLCC (ultra-reliable and low latency communications) has requirements of a single millisecond to traverse, one way, the entire network infrastructure. This is required to enable manufacturing and robotic control, and safety critical use cases. Therefore, without the latency requirement being met, not only are costs high, but the fundamental premise of 5G (enabling new vertical use cases) is damaged.
Geo-distribution of a single cloud infrastructure
The geo-distribution of the cloud infrastructure is not optional in 5G, it is a physical reality every service provider faces. Typically, service providers cover thousands to millions of square miles of territory, and the RU element must be placed across that territory to deploy 5G. When vRAN is introduced, the vDU and vCU functions likewise become distributed across the physical topology.
Hence, the geo-distributed cloud is created. With this comes the necessity of supporting this as a single managed entity, to ensure the operational aspects of deploying and operating it are realistic. If the selected solution requires the operator’s personnel to physically visit, or even uniquely remotely access each site through separate systems, then the operational burden of the 5G network is enormous. Many cloud solutions are incapable of providing a single pane of glass management solution with zero-touch automation for software upgrade and deployment across the cloud infrastructure.
The future is bright for 5G
With Wind River Cloud Platform, a fully open source solution based on the cloud native project StarlingX, service providers worldwide finally have a true option for deploying 5G vRAN. Now that a geo-distributed cloud may be deployed and maintained in a cost-effective manner, meeting the key requirements of vRAN technology and applications, the future is certainly bright for 5G.