Multi-G Initiative forms to speed up software deployment within Open RAN platforms

Smoothing the path to market for a clever MU-MIMO algorithm, or radically overhauling a key underlying platform for Open RAN. Or both?

A group of companies, led by RAN software developer Cohere Technologies, are convening as a group to enable more flexible interworking between Intel’s FlexRAN platform and apps that operate radio control functions in the Open RAN RIC (Radio Intelligent Controller).

A grand-sounding press release said the work the companies take on will: 

  • Pave the way for large scale deployments of Open Radio Access Networks
  • Develop the industry’s first framework for a multi-generational (Multi-G), software-based Open RAN architecture
  • Define frameworks, interfaces, interoperability testing, and evaluation criteria that would provide the interfaces to support full coexistence of 4G, 5G, and future waveforms

It would do this by “disaggregating RAN intelligence and scheduling functions, enabling future code releases of Intel’s FlexRAN reference architecture to support higher capacity, software-defined deployments for 4G, 5G and next generation wireless waveforms and standards.”

They have called this group the Multi-G Initiative and said that they will hold their first meeting in May this year. The members are Cohere Technologies, Intel, Juniper Networks, Mavenir and VMWare.

So what is going on? Is this the start of a radical opening up of the platforms that underpin an Open vRAN – extracting intelligence away from the underlying software and hardware platforms into platforms such as the RIC? Or is it about something more narrowly commercial, more limited to the interests of the companies involved – i.e. a group of interested parties bringing Cohere Technologies’ capabilities to market?

In the first instance, it does sound like the root of the work is being commercially driven by the desire of the parties involved to fully exploit capabilities developed by Cohere Technologies and by RIC providers. Cohere has been marketing its universal Spectrum Multiplier (USM) software – a means of enabling spatial multiplexing MU-MIMO. It has been doing so by delivering the algorithm as an xApp on a near real time RIC, where it performs the channel calculations necessary to enable users on the same spectrum to share the same space.

Vodafone, notably, as well as Bell Canada and Telstra (both are Cohere investors) have been trialing the software, which Cohere says can double spectrum efficiency. So far it has achieved good reports from its trial operators. 

Cohere has calculated that its best route to market is within the Open RAN ecosystem. In pilots, it has deployed USM as an xApp on RICs developed by Juniper and VMWare, and with support from vRAN vendor Mavenir. Mavenir’s vRAN DU-CU baseband products have run, to date, on Intel’s FlexRAN platform. FlexRAN is a L1 software reference platform that is tied to Intel’s chipset and accelerators that sit inside the hardware that operators the vRAN workloads. For a vRAN provider like Mavenir, being able to integrate with and support an app like Cohere’s could give it a working commercal advantage over other vRAN providers that have not done the work.

To carry out these spatial multiplexing calculations for all “Gs”, including a putative future 6G waveform, the companies have realised that Intel will need to make some changes to its FlexRAN platform. There are also changes required to the interface between the RIC (Open RAN Radio Intelligent Controller) and Layer 1 in the radio.

In the words emailed by its CMO Ronny Haraldsvik to TMN, Cohere said, “This requires a redesign of the various control functions that are embedded in each generation, moving them to a common set of controls which are embedded in Cohere’s solution, and leaving behind native Layer 1’s. This is a huge step, and will make our solution work transparently across 4G/5G, and it will make adding OTFS/6G easier down the road.  This also makes our integration with Mavenir (and future RAN partners) easier as we will meet at standardised interfaces.”

I think this marks the start of making FlexRAN more open and more powerful

Santiago Tenorio, Vodafone Fellow and Director of Network Architecture, and Chairman of the Telecom Infra Project (TIP), said in written comments to TMN, “USM requires both specific support from L1 (currently Intel, later others may be willing to support the changes too) as well as some further standardisation in the RIC space.”

Take a look at both these statements and ponder. If “a common set of control functions” were moved into the RIC space and exploited by solutions such as Cohere’s, thereby leaving behind a “native” Layer 1, are we also seeing a move that would see Intel’s L1 being more open – reducing the dependencies between its FlexRAN software platform and its underlying chip and hardware accelerators?

This perceived dependency – that once a vRAN provider and operator are reliant on Intel-based servers and FlexRAN, it is very hard to port – has been a bone of contention for some in the industry, including Tenorio. It ties a software provider to a hardware platform, which runs counter to the disaggregated notion of an Open RAN. Rakuten Symphony’s CEO Tareq Amin is another that has suggested the industry would benefit from a more open L1 software environment: “I think there needs to be a reference design that gets created for everyone“.

As for the goals of the Multi-G Initiative itself, Vodafone’s Tenorio said, “I think this marks the start of making FlexRAN more open and more powerful.” And he adds, “Nothing prevents other L1’s to do the same and enable Cohere. Intel has just been first, same as Mavenir. But others should and will come.”

It’s worth noting also Tenorio’s prepared comments in the press release, in which he said the commitment towards a software programmable L1 stack, is fully aligned with the vision of Open RAN and will bring us one step closer to the scale deployment of software-defined RAN.

Intel*, for its part, rather indicated that FlexRAN, being already fully software programmable, can enable the Multi-G framework. In comments attributed to Sachin Katti, SVP & GM of the Network and Edge Group at Intel Corporation, it said, “This Multi-G framework, enabled by Intel FlexRAN – which is fully software programmable down to L1 – will enable faster O-RAN adoption and unlock new innovations.”

* UPDATE: Intel provided the following additional quote in response to a request from TMN:

“For years, FlexRAN has powered vRAN to meet the choice and flexibility of our customers’ vRAN requirements and solutions. We have a wide and vibrant ecosystem and we will continue evolving FlexRAN to deliver the best outcomes for our customers. At this time we have nothing further to share yet we look forward to providing updates as the Mult-5G initiative convenes in the coming months.” – Cristina Rodriguez, Vice President, Network & Edge Group and General Manager, Wireless Access Network Division (WAND).

In the long term, companies know that the level of investment that operators sunk into 4G and 5G upgrades may not be sustainable.

Haraldsvik referred TMN to this LinkedIn post by Rakuten’s Amin, which noted Omdia‘s work on telco capex, and the large chunk accounted for by the 5G RAN.

So, we asked, “Is the Multi G Initiative about stopping the G? So you don’t have to re-deploy hardware each time there’s a new G – you can do new scheduling/waveform in software with interface to the L1?”

“Yes,” Haraldsvik replied. “You are dead on.”