Virtual RAN: Separating Fact from Fiction

Agility and flexibility rather than “one size fits all”

It’s hard to escape the hype for 5G, but there’s another area where industry attention is now almost as palpable: the radio access network (RAN). This is a considerable change given that the RAN has operated for years as a highly distributed and proprietary part of the network. However, the need for more flexibility, agility and lower costs is pushing operators to think about new and different deployment options for the RAN.

One route that many operators are taking is to virtualize the RAN and move closer in design to the architecture that has underpinned the growth of cloud platforms. Rather than the network being made up of a wide array of fixed-function devices, network capabilities are virtualized and run on standardized hardware. The hardware is abstracted away through software in order to cut costs and increase agility and efficiency.

This means network functions can be run in a container or a virtual machine in much the same way as computing functions run in data centres. Network functions are programmed in software, making it possible to use commercial off-the-shelf hardware and implement tools that were previously tied to specific hardware. The result is optimization and easy redistribution depending on the needs of the network or applications, as well as improved ability to support new services and uses.

For 5G to deliver its promise, this broader transformation of the RAN and core network is essential. In the near term, lower latency, better uplink performance and greater efficiency will all need a shift to a dedicated 5G RAN. Indeed, the dramatic change in network loads and traffic caused by the shift to home working during the Covid-19 pandemic is a perfect illustration of the need for a more flexible and adaptable network. Operators must be able to dynamically control network capacity and prioritize traffic, particularly as 5G enables a raft of mission-critical applications.

However, this common agreement about the merits of virtual RAN also oversimplifies how operators are adapting and transforming their networks. Not all operators are the same, with each facing different competitive pressures, priorities and customer requirements. Moreover, network architectures and spectrum assets vary wildly. Nonetheless, the exponential growth in data traffic and the need to ensure capacity form a common dynamic affecting all players.

This is why we believe operators will adopt a variety of approaches in the transformation of the RAN and why it almost certainly won’t be a case of “one size fits all”. Open RAN in particular is grabbing a lot of attention as it seeks to standardize the design and functionality of hardware and software elements. It’s an important enabler of the “cloudification” of networks. Virtualization, however, is not a new concept.

There are also counter-arguments to Open RAN specifically. Common criticisms of the approach are the identification of responsibility amid a sea of suppliers, higher operating and maintenance costs and the challenge of testing and verification. Equally, the Open RAN camp levels similar arguments against the established approaches of proprietary hardware and software. Although the relative immaturity of Open RAN means its “plug and play” vision is some way off, the momentum is promising.

We’ve spoken with a host of operators and major ecosystem players in recent weeks and found a wide variety of perspectives. This points to a continuum that starts with those pursuing a more traditional and decentralized approach, and ranges to the likes of Rakuten, which is embracing a fully centralized and virtualized network. In between these two opposites lie a host of operators at varying points of maturity.

Rakuten is the most advanced on this journey and claims to be the first operator to deploy a fully virtualized network. It plans to use Altiostar’s software to separate 5G network functions into containerized applications that will run on servers based on Intel Xeon Scalable processors. These applications will use Kubernetes plug-ins that enable RAN and mobile edge computing containers to be deployed on Intel architecture.

Rakuten has the unique advantage of being a greenfield network without the restrictions of a legacy architecture, but it’s no by no means alone on the path to virtualization. Verizon claims it is now virtualized from its core network to the edge, and it’s joined by a flurry of operators that have made recent announcements. In an interview with CCS Insight, Santiago Tenorio, head of network strategy and architecture for Vodafone, said that the operator expects to make the first commercial deployments of Open RAN within the next two years and possibly sooner.

Telefonica announced in March 2020 that it would further an existing trial of Open RAN on 4G by extending it to 5G in Brazil, Germany, Spain and the UK in 2020. In conjunction with Altiostar, Gigatera Communications, Intel, Supermicro and Xilinx, the trial will encompass the planning, design, testing, integration and deployment of disaggregated RAN systems using baseband radio functions running on Intel’s Xeon servers using its FlexRAN reference platform. Similarly, Deutsche Telekom recently announced that it’s teaming with VMware and Intel to create an intelligent virtual RAN platform built on Intel’s FlexRAN architecture and running virtual RAN workloads on VMware’s Telco Cloud platform.

The common thread among the different approaches being taken to virtual RAN and various phases of commercialization is the increase in processing requirements and the importance of partners that can deliver at scale while embracing and contributing to open standards and software development. Virtual RAN is closely following the playbook that established the cloud, and it’s notable that Intel is playing a central role in many deployments across hardware and software.

Operators’ future revenues depend on success with 5G, so they have a vested interest in accelerating the transformation of their networks. If 5G is to achieve its promise and deliver rich capabilities at the network edge, a high-capacity and low-latency air interface isn’t enough: it must be supported by a network architecture that has the flexibility and agility to match.

A version of this blog post was first published by FierceWireless on 11 May 2020.