The Future of RAN is Virtualized and Open

The global appetite for mobile data and services is growing rapidly as the digital revolution continues. Smartphones have become an indispensable part of our daily lives. We use them to take photos, get news and social updates, entertain ourselves, hail and track ride-sharing services, or maybe play a game – and of course occasionally make a phone call. In addition, the magic of software and “app stores” means that we have become used to having our smartphones deliver new and innovative user experiences without upgrading the device itself. We simply download or update an app.

Could our network infrastructure be similarly software-defined? What if wireless networks – the ones that connect your mobile device to the apps and services you use daily – could be upgraded in the same way with a simple software update? Unbelievable? It’s not, and Intel is fully committed – and leading the ecosystem – to this future.

For more than a decade, Intel has been leading network transformation with its ecosystem partners and customers to make this vision a reality. As the leading 4G and 5G network silicon provider, it has been our mission to transform and liberate these networks to become “virtualized” and to run on software. This gives communication service providers (CoSPs) the programmability and upgradability in their networks to deliver new capabilities and new services – just like you get through a simple software update or new app download on your smartphone or PC.

And we’re seeing the fruits of this labor finally pay off.

Take the brains of the network, called the core: 10-plus years ago, there were barely any wireless networks that could run the core as a piece of software on general-purpose programmable servers. The conventional wisdom was that it was impossible to make the core network software-defined. Intel took that challenge; we designed servers and software that helped move the core of the network to be virtualized and run as software on general-purpose servers. In 2020, such virtualized core network deployments accounted for about half of all deployments, delivering on the software-defined network vision. And these virtualized core deployments are forecasted, according to a report by Dell’Oro, to reach over 90% of core network deployments by the end of this year¹, with almost all known virtualized network servers running on Intel CPUs.

The radio access network (RAN) is the other half of the network, and in fact is the part of the network our phones communicate with directly when we make a call or browse the web or use an app. Similar to core networks a decade ago, the conventional wisdom is that it will be very challenging to make the RAN software-defined; today most of the world’s wireless networks implement the RAN as fixed-function appliances that cannot be changed once deployed. Intel’s ambition is to drive the same transformation in the RAN that we drove in the core network a decade ago. From customer co-investments to cutting-edge technology, Intel is committed to helping operators and OEMs bring to market virtualized RAN as a new and lower-cost alternative to high-speed communications networking.

Virtualizing the RAN is daunting. It is a very demanding piece of the network infrastructure and the cornerstone to ensure your most important call is never dropped. It needs to deliver the high-speed and low-latency connectivity we all desire, but without ever failing, since we literally depend on it during emergencies. It is also the most expensive piece of the wireless network: More than half of the budget to build a wireless network is spent on the RAN. Designing and delivering the hardware – and software – with the right combination of performance, reliability and cost-effectiveness needed to meet the RAN’s demanding requirements is no easy task – technically or financially.

The Answer: Full Flexibility, Less Complexity

The costs to replace outdated hardware when new requirements, new use cases or new standards hit the market can be unfeasible. CoSPs, like any business, want to get the most from their investments. They do that by reducing total cost of ownership, saving power and creating platforms for continuous innovation.

Network modernization to software-defined infrastructure does just that: It makes upgrades simple, helps component reduction requirements, and reduces system and board complexity and bill-of-materials costs, while enhancing supply chain diversity. And software allows for programmability, offering the ability to innovate and iteratively improve the network through software upgrades, which significantly improves hardware deployments’ return on investment.

Traditionally, RAN infrastructure has run on custom fixed-function hardware. In other words, they are not software-defined; upgrading them usually meant expensive and cumbersome hardware upgrades. The reason is that there is a belief that the lowest layer of the RAN architecture – Layer 1, or the part that translates the radio signals from our smartphones into bits and delivers the high-speed 5G connectivity we all enjoy – is so demanding in its performance requirements that it cannot be virtualized and defined in software.

That’s where Intel architecture comes in. We believe that virtualizing the RAN – all the way through to Layer 1 in the RAN software stack, where the digital data is transmitted – is necessary to deliver on the benefits of virtualization. To handle the demanding needs of the RAN, including Layer 1, we have built an architecture based on a flexible, programmable general-purpose chip, integrated with acceleration for the most demanding tasks, that enables CoSPs to deploy a fully virtualized RAN without compromises and reap the full benefits of having a software-defined network end to end, both the core and the RAN.

It’s why we continue to innovate – to make communications networks work faster and better, and to deliver services we all now expect. Our general-purpose network chips are constantly evolving, and the latest 4th Gen Intel® Xeon® Scalable processors with Intel® vRAN Boost are poised to launch at the biggest connectivity show in mobile networking: MWC Barcelona 2023, alongside some of our largest customers.

General-Purpose Chips are the Backbone to Virtualization

Full virtualization of the RAN, all the way through to Layer 1, offers immense technical and business benefits, including agility, flexibility and scalability. It supports new innovations, like AI algorithms across functions in the RAN, evolving networks to deliver ever-more capabilities at an optimized cost. While some of these benefits may be hard to notice directly, if you are reading this article on a mobile device, you’re likely connected to virtualized network infrastructure running on Intel.

Operators can easily implement dynamic power management and network function redistribution. Failures in network operation and system upgrades can be handled by moving the network workload to a different server – without sending technicians to the field. With our general-purpose chips, operators can easily turn cores off to save power in times of low load – or no load.

On the other hand, having a Layer 1 accelerator card means that reconfiguring existing networks to support new RAN technologies and services is expensive. It requires new hardware and associated costs, and, ultimately, will be affected by a scarcity of talent able to build products with these devices.

With much of the Layer 1 processing completely offloaded to a custom chip, custom software-based functions of Layer 1 of the RAN stack are executed (and often hand-coded) in proprietary software languages. These Layer 1 accelerator cards also rely on proprietary tools for compiling, debugging and building applications.

Simply put, having custom silicon in your RAN means it’s not virtualized.

This is not the case for general-purpose chips. Within software-based architectures, to integrate – or get the most out of general-purpose hardware – the software is written in standard open programming languages that leverage generic compiling, debugging and build tools. The benefit? Software written for one chip generation is reusable in follow-on generations and easily ported from one generation to another, allowing operators to consolidate the RAN software on a common virtualization platform.

In other words: Write once, deploy everywhere. Industry leaders who offer data connectivity to the world agree that general-purpose chips meet their challenges and demands.

A New Approach to 5G Infrastructure

A true disaggregation between hardware and software lets operators buy best-in-class hardware and software components from different vendors, thus enabling vendor choice at a very granular level. And who doesn’t like choice?

4th Gen Intel Xeon with Intel vRAN Boost chips do exactly what the name says: They boost networks by offering a programmable infrastructure and eliminating the need for custom Layer 1 accelerator cards. This new chip integrates vRAN acceleration directly into the Intel Xeon system-on-chip. It is natively designed to power cloud-ready virtualized RANs and it will deliver up to twice the capacity within the same power envelope² and an additional 20% power savings due to integrated acceleration³ to address critical performance, scaling and energy efficiency needs for operators.

We’ve chosen an integrated acceleration approach because it combines the benefits of inline acceleration with the flexibility and programmability of x86, and that is preferable to any solution that shoves an entire Layer 1 into an inflexible hardware accelerator. And our 4th Gen Xeon Scalable Processors with Intel vRAN Boost are expected to match or better the performance-per-watt of the best custom dedicated Layer 1 accelerator cards when they enter the market⁴.

vRAN is the Future

Years of research and development, ecosystem collaboration, commercial deployments, lessons from successes and customer feedback tell us one thing: Networks need to move to software. To do so, the wireless industry needs a solution to virtualize Layer1.

And as networking has been going through a long transformation to become software-defined, Intel’s vRAN products and FlexRAN reference software have evolved to meet this moment. Intel has powered this transformation from infrastructure built with fixed-function hardware to fully virtualized platforms running on commercial off-the-shelf hardware based on our general-purpose processors.

Intel’s is a deployment-ready solution with a future roadmap that is first-to-market, and this instills confidence in operators. They can adopt virtualization knowing that Intel not only provides the solution today but will continue to evolve the solution – as people come up with even more ways to use their “phones.” That is why nearly all of today’s commercial vRAN networks around the world run on Intel architecture.

More Context: For more on the subject, read “Benefits of Virtualizing the Layer 1 in a RAN Stack,” an Intel white paper.

Sachin Katti is senior vice president and general manager of the Network and Edge Group at Intel Corporation.

Editor’s Note: The description of commercial vRAN networks running on Intel architecture was edited on Feb. 8, 2023.

_{¹ Dell’Oro report published in Jan. 2023, “Mobile Core Network & Multi-Access Edge Computing Quarterly Report” + Intel internal analysis
² Estimated as of 8/30/2022 based on 4th Generation Intel® Xeon® Scalable processor architecture improvements vs. 3rd Generation Intel® Xeon®  Scalable processor at similar core count, socket power, and frequency on a test scenario using FlexRAN™ software. Results may vary. Performance varies by use, configuration, and other factors. Learn more at www.Intel.com/PerformanceIndex
³ Estimated as of 08/30/2022 based on scenario design power (SDP) analysis on pre-production 4th Generation Intel® Xeon®  Scalable processor with Intel® vRAN Boost (integrated vRAN acceleration) and pre-production 4th Generation Intel® Xeon®  Scalable processor at the same core count and frequency with an external vRAN accelerator card. Performance and power vary by use, configuration, and other factors. Learn more at www.Intel.com/PerformanceIndex
⁴ Performance/power projections are based on Intel estimates and simulations as of October 2022.}