Virtualized 5G RAN: why, when and how?

Virtualization of radio access network (RAN) functionality is a hot topic, and we see a rapidly growing interest in understanding what the impact and benefits will be for service providers when deploying 5G.

network operations center

Head of Product Line RAN Compute

Head of New Architectures within Product Area Networks

Head of Product Line RAN Compute

Contributor (+1)

Head of New Architectures within Product Area Networks

Category & Hashtags

Virtualized radio access networks in a nutshell

Virtualization entails the migration from custom-built network nodes to network functionality implemented in software running on a generic hardware compute platform.

Virtualization for communications service providers began with the core network and subsequently cloud technologies have been evolving at a rapid rate.  In the RAN domain, vendor agnostic commercial off-the-shelf (COTS) hardware has the potential to enable innovation across a range of software ecosystems.  

A separation of the upper and lower parts of the RAN was standardized in 3GPP R15, where a higher-layer split was specified with a well-defined interface (F1) between two logical units: the Centralized Unit (CU) and the Distributed Unit (DU). The CU—with less stringent processing requirements—has been more amenable to virtualization than the DU and its functions that are closer to the radio.

For full-stack RAN virtualization, the DU is connected to the radio via a packet interface known as enhanced Common Public Radio Interface (eCPRI). There are multiple ways to divide functions between the DU and the radio; in standards discussions these are referred to as lower-layer split (LLS) options. One possible alternative specified by the O-RAN Alliance is referred to as the 7-2x split, but other functional splits are also being considered.

Benefits with virtualized RAN

RAN virtualization presents several significant challenges as the processing and timing requirements are very high on certain functions in the lower parts of the RAN. These functions are critical as many aspects of RAN capacity and coverage are determined by them. Still, the potential benefits that virtualization could bring may very well be significant.

First, a fully virtualized RAN (vRAN) could bring significant benefits of harmonization: one single uniform hardware platform across the core network, RAN and edge. This could simplify the management of the complete network, reducing operations and maintenance costs.

Second, in a full vRAN, the network functions will be separated from the processing hardware. This means that RAN network functions from multiple vendors could run on the same hardware, increasing the flexibility for the service provider.  In some cases, the hardware could even be shared between service providers.

Third, vRAN offers an opportunity to embrace established solutions, available in today’s public cloud technologies, for non-RAN-specific functions. By agreeing to use industry-established components for common tasks, the need for costly adaptations of vendor-specific solutions would be removed. If this is achieved, it would allow the RAN ecosystem to focus on business-critical components.

Fourth, a vRAN holds the promise of increased flexibility as functionality and capacity could be more easily deployed where and when required. Cloud technologies could facilitate this type of flexibility.

Finally, a widely adopted open platform could also lower barriers for cross-domain innovation, facilitating the development of new use cases and services.

Benefits with a purpose-built RAN

In comparing vRAN with the kind of prepackaged, pre-integrated end-to-end solution that is delivered today in the form of the Ericsson Radio System, there are several points to bear in mind.

First, purpose-built RAN hardware will most likely always provide a higher level of efficiency in terms of size and power consumption. This is due to the optimization of network functions that are implemented in custom hardware rather than on generic processors.

Second, for delivering RAN functionality the cost of custom-built hardware will be lower compared to vRAN. However, over its lifetime, a COTS-based platform could become increasingly cost efficient due to hardware manufacturing volumes across multiple industries.

Third, the purpose-built RAN solution is a verified and pre-integrated end-to-end system, whereas in many cases, a virtualized system—if not procured from a single vendor—will require additional system integration efforts and costs that need to be considered.

Fourth, in a pre-integrated end-to-end solution from a single supplier such as Ericsson Radio System, the system performance accountability is clear. With a virtualized solution—comprised of hardware, software and service elements from multiple suppliers—accountability needs to be clearly assigned.

The question remains if and when the potential operational benefits and flexibility offered by vRAN can outweigh and compensate for the hardware, power and system integration costs, while attaining the same high level of system network performance. One key aspect will be the cost evolution of custom-built hardware compared with generic COTS hardware and the emerging COTS-based accelerated compute platform.

Addressing 5G use cases

The evolution to 5G networks is the result of continuous improvement of telecommunications technologies by the 3GPP partnership. Each new release has brought enhanced capabilities including supporting more spectrum, additional frequency bands, as well as air interface enhancements in performance and efficiency. One consequence of this evolution is that the processing requirement for the network functions also increases. With today’s radio access technology, RAN workloads can run on a general-purpose computing architecture based on processors such as x86 central processing units (CPUs)—but for full 5G capabilities more processing power will be needed. One approach is the addition of NVIDIA graphics processing units (GPUs), with which the CPU can be offloaded. The combined CPU and GPU-accelerated platform should have the potential to handle even the most demanding 5G use cases. In order to realize the benefits of vRAN, the platform must be open and based on COTS hardware that is fully adopted by the cloud ecosystem with a large developer community supporting it.

Collaboration between NVIDIA and Ericsson

Ericsson is collaborating with NVIDIA—the leader in accelerated computing technologies—to build high performance virtualized 5G radio access networks using NVIDIA’s GPUs and software.

The aim of the collaboration is to establish an open platform design – not proprietary to Ericsson, NVIDIA or the CPU vendor. Ecosystem alignment will be needed, so server offerings can be standardized. The goal for Ericsson will be to commercialize vRAN technologies through a new product portfolio to deliver radio networks with the flexibility and agility to meet tomorrow’s market requirements.

Open interfaces

There is work ongoing by the O-RAN Alliance – to which Ericsson is an active contributor – on the development of specifications for open network interfaces. The latest approved version of the O-RAN document can be found here. (To download a copy of the O-RAN Specification, you must agree to the O-RAN Adopter License Agreement.)

Virtualization and open RAN interfaces as defined by 3GPP and the O-RAN Alliance are often discussed in the same context, although they are mutually independent. From a technical perspective, open interfaces can be realized on both traditional purpose-built and new COTS-based platforms.

Regardless of how virtualization is achieved, as the hardware and software are not necessarily aggregated by a network vendor, these systems will necessitate new business models with clearly defined accountability. System integration, whether managed by a RAN software supplier or a cloud infrastructure provider, will be crucial to ensure network performance and reliability. Regardless which party takes on this responsibility, the additional costs due to increased system integration work cannot be overlooked.

Ericsson future-proofing its RAN portfolio

Even if most of the technology components required to build high-performance virtualized RAN products are in place, the main question remains - when will virtualized RAN be a viable business proposition? Independent of when, Ericsson’s main objective is to provide future technologies and products that fulfill our customers’ needs and requirements. Ericsson is actively working – independently and within the ecosystem – to explore COTS-based platforms that bring higher efficiency, and also with harmonization and standardization of vRAN components and architectures to minimize the cost of system integration. The target of this work is to make vRAN a commercially competitive solution that can complement existing investments in the near future.

Exploring how to build a high performing vRAN solution is one of several initiatives Ericsson is pursuing to ensure a successful 5G market. We are on an exciting journey and the future for 5G looks bright.

This is the first of two article on 5G RAN virtualization. The second, entitled “Accelerating 5G vRAN: The Case for GPUs” can be found here.

The Ericsson Blog

Like what you’re reading? Please sign up for email updates on your favorite topics.

Subscribe now

At the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.