Nokia announces Open RAN 5G products

Company announces upcoming availability of new vRAN elements for 5G.

Nokia has announced the upcoming release of an updated version of its virtualised 5G Airscale cloud RAN portfolio. New 5G vRAN functions will be compatible with O-RAN interface specifications, it said, meaning that Nokia becomes one of the first companies, and the first “Big 3” vendor, to take the wraps off an Open RAN 5G product line.

The5G AirScale Cloud RAN solution based on vRAN2.0″, will be “commercially available” this year but on general release in 2021, Nokia said.

The vRAN 2.0 release adds a virtual Distributed Unit (vDU) to the existing virtual Central Unit (vCU) that has been in commercial use since 2019. Nokia has also developed a fronthaul gateway that can be used either with Nokia radios or with those of other vendors. The vDU and the fronthaul gateway can, like the vCU, run on Nokia’s x86-based AirFrame Open Edge Servers and AirFrame Rackmount Servers.

Interfaces between the split baseband elements (vCU/vDU) and between the vDU and the remote radio will support O-RAN specifications, the vendor said. That means that in theory operators could buy “all Nokia” Airscale and remote radio and antenna unit solutions, or mix vendors across the different domains. 

The vDU is using Intel’s FPGA-based Vista Creek hardware acceleration for parts of the real-time sensitive L1 computing in the vDU. Nokia said it would have integrated GPU-based acceleration in the vDU by 2022 – a development it described as aligning with vRAN 3.0.

Michael Clever, Senior Vice President and Head of Cloud RAN Edge Cloud Platforms, Mobile Networks, Nokia, said that the use of FPGA based accelerator cards for Layer 1 processing such as FEC (Forward Error Correction) or LDPC (Low Density Parity Check) coding is sufficient to get the job done and,  importantly, products are available now. 

Clever said, “GPU’s still need some significant development work to adapt them to this specific L1 use case. However they are expected to be able to provide additional capabilities; for example in combination with AI/ML based features and services. But this will take some time, so we expect at least until 2022 to be dominated by FPGA based solutions for accelerator cards.”

Ericsson has also announced it is working on GPU-based acceleration for lower layer vRAN use cases, in partnership in its case with Nvidia.

Nokia announced earlier this year it would be working with Marvell to include its ARM-based multicore designs within the ReefShark 5G chipset family that powers its Airscale products. It has also more recently said that it will use SoC products from Broadcom. The company is building out a selection of options for deployment across its Airscale designs. 

Architecture options

Nokia said that the vRAN2.0 architecture options meant that the vDU and vCU could be co-hosted at the same central location, or split between central and edge sites, or the vCU could be at an edge site with the vDU integrated directly onto the antenna site. This last integration is known as a Radio Access Point (RAP).  

Clever said that a centralised implementation is “usually based on dark fibre links between the radio sites and the centralised location. There might be also some Fronthaul Switches or Fronthaul Gateways involved for traffic concentration or conversion to legacy radios.”

“The architectural deployment is a “trade off” between transport costs (lower in distributed configurations), service costs (lower in centralised configurations) and pooling gains (higher in centralsed deployments). That’s why we support all options,” Clever said.

Nokia’s press release said that its AirFrame open Edge Server is “the first x86-based product in the world built and tailored for edge cloud and far edge cloud deployments of Cloud RAN.” 

That may be a moot point with, for example, Altiostar, which has edge implementations of Cloud RAN in Japan on x86 servers.  Clever said that it is correct that Altiostar is also using an X86 based edge implementation with FPGA based accelerator cards, but he added that Open Edge servers are “specifically designed for far edge deployments, including all the specific footprint, performance, power and heat requirements. It has also been commercially already since 2019.”