History, it is said, is written by the victors. If we go by the historiography on offer this week at FYUZ, you could think that Ericsson, and AT&T, are the winners in Open RAN.
Certainly one theme within the Open RAN track at TIP was the prominence given to the path AT&T has taken with Ericsson.
AT&T’s Rob Soni is now TIP chair, and he spoke multiple times across the event, both with his AT&T and TIP hats on (in fact he wore an AT&T teeshirt at one point to make the distinction).
With his TIP (invisible) hat on, Soni said that progress being made by big US operators will commercialise Open RAN solutions to the benefit of all. AT&T’s work to insist on open interfaces with Ericsson, and integrate with a third party RU vendor (Fujitsu), shows a path forward to mass commercialisation. He also said he expected his predecessor as TIP chair, Yago Tenorio,who recently left Vodafone for Verizon, to drive his new employer further towards Open RAN and Open RAN technology. (Verizon, by the way, already has a widescale vRAN deployment with Samsung, Wind River and others in the stack.)
In what could only be classed as the mildest of digs, he said that Dish (or Boost Mobile) is also now gaining scale “due to some nice financing”. Perhaps only the most cynical would consider this a somewhat ungenerous portrayal of a company that has rolled out at the speed of its licence requirements following a multi-vendor template: horizontally from CU-DU to RU and vertically through the hardware to software stack, including within public cloud locations.
But the point was, large commercial deployments of Open RAN, within the definitions provided, are happening. And this solves issues for the rest.
And while large operators are driving this, there’s a gap underneath, Soni said, where there is an opportunity for “disruptors” to insert themselves in the network. Once the issues of integration at scale are worked out, then smaller operators will be able to roll out without the complexity and challenges that a lot of operators are currently facing.
And those challenges are not about cost, per se. More, they are about lifecycle management of equipment and software, being able to deploy cloud-based resources where and when you need them (core, edge, cell site) at the scale you need them, in processes that are divorced from the standards-based generational advances. And yet, it’s not clear how the Ericsson/AT&T path is one that helps operators integrate “disruptors” in other operators’ networks.
So it was interesting then to hear Soni, very much with an AT&T teeshirt on, expound on the adoption of Ericsson as the cure for Open RAN integration. The way he told it, in around 2022 the O-RAN Alliance was at an impasse and unable to agree on how to deal with uplink in the Open Fronthaul interface for higher order MIMO. In fact in widely reported comments he said the O-RAN Alliance was “out to lunch”. The first thing Soni was told to do at AT&T, he said, was to fix it.
The solution, as we know, was Ericsson’s proposal for Uplink Performance Improvement, which put more time-sensitive processing in the RU, reducing the performance requirement across the fronthaul link from DU to RU. What it required, though, was RUs with the necessary hardware to support that, plus of course a tweak on the DU. Ericsson’s later model RUs can, with a software upgrade, handle that processing requirement. In fact it said it had a million ULPI-capable RUs in the field already.
And yet, what Soni didn’t mention was that there was indeed a MIMO category already available – Cat B.
Cat B already existed and was O-RAN Alliance’s defined standard for massive MIMO. It was supported by the likes of Samsung, as well as newcomers Mavenir and Fujitsu. In fact, to be truly compatible with O-RAN Alliance standards a vendor should support all categories in a backwards compatible manner, which means if you support Cat B ULPI you shoudl also be able to support Cat B. Some, like Mavenir, which was no fan of the ULPI proposal, have recognised market requirements and will support all the interfaces. Ericsson does not do this. It’s just that Ericsson RUs were not Cat B compatible, and Ericsson didn’t want them to be. It argued, with some backing from operators (including Vodafone), that Cat B without ULPI was inefficient.
And of course one of those operators backing the new proposal was AT&T, which not only had a lot of Ericsson Radio Units in its network that were not and woudldnot be Cat B compatible, but was also about to hand over even more of its network to Ericsson. And that’s why the two Open RAN winners want to position what they have done as solving Open RAN interopability.
Cloud RAN
But history lessons aside, we should also look forward. One aspect that Soni mentioned that is worth mentioning is the operator’s desire to move to more deployments of Ericsson’s Cloud RAN platform as part of its three year modernisation programme (actually Soni initially said replacement programme and then corrected himself) that will involve 70,000 sites.
The operator had learnt a lot already on how to introduce cloud in a distributed way. Cloudification, Soni said, is an important part of managing portability, so it is not locked into any dimension.
By moving Ericsson from being a vertically integrated to a horizontal player, the operator will be able to create a new network programmable layer. It can migrate legacy apps or tooling to one interface, rather than having three SON engines each with its own interface and API.
“We have largely seen automation as per domain. With vendor-specific tooling and scripts. Now we can take action in a common way and with that common interface leverage AI techniques.”
But maybe there’s something else happening here. Ericsson said in a blog post this week that it is indeed adding support for Cat B ULPI to its Cloud RAN platform from 2025. It added that its purpose-built RAN and its Cloud RAN will support Cat A (which is for lower order MIMO) from 2025. According to reporting from Light Reading it has confirmed that it would not be supporting Cat B ULPI from its purpose-built RAN. This is a departure from Ericsson’s previous messaging, where it said that its whole aim was to proceed with parity between its purpose-built and Cloud RAN compute platforms. It explicitly said both would move to support ULPI as the fronthaul link of choice.
In this article from March 2023, we quoted Marten Lerner, Head of Cloud RAN Product Line, Ericsson, stating “When we look at this from a long term vision, we see that the NG LLS will be the interface that we will use for our radios across our portfolio going forward.” NG LLS was Ericsson’s internal name for what became Cat B ULPI.
It’s still possibile, of course, that the purpose-built platform will migrate. A “long term vision” can be just that. And at the time Lerner said:
“So this is will of course not happen from day one, we have a lot of installed base that we need to take care of and all of that. But the long term vision is that the interface we will have between the compute side, and the radios will be NG LLS.
“So this is our way of saying that we are not we’re not half in on LLS , we’re not doing that as some innovation track on the side. We are now going all in on making sure that this is the long term portfolio of Ericsson based on the splits that has been defined.”
This image from 2022 Capital Markets day is also explicit. Both the ERS Ran Compute (purpose built) and the Cloud RAN platform will interface to the radio over NG LLS (ie. Cat B ULPI).
If that turns out not to be the case then either Ericsson backers (AT&T is not alone, Telefonica and Orange in Spain have backed the Ericsson path) will notbe able to have true Open RAN to their massive MIMO radios or they will need to migrate to Cloud RAN in order to do so.
Small Cells
Was AT&T’s announcement of Fujitsu as a confirmed third party supplier of small cell radio units within its network quite the validation of its commitment to a multi-verse that it was said to be?
The carrier has 100,000 sites, and 70 percent of those will be replaced (or modernised as Soni corrected himself) by the terms of its new Ericsson deal. Meanwhile its announcement of Fujitsu is to provide its small cells as a third party RU in select use cases, under the AT&T shroud. There were no indications of volume or scale. Meanwhile the same Ericsson blogpost as contained the “no ULPI for purpose built RAN” nugget, said that third party RU testing was in essence just beginning, under the auspices of the ACCORD lab.
OTHER OPERATORS AT FYUZ
Vodafone isn’t rowing back
Other operators at the event were still, mostly, grappling with the implications of integrating and then rolling out at scale disaggregated, open radio networks.
A year ago, Vodafone was reiterating its commitment to dedicate 30% of its 170,000 site tender for Open RAN compliant vendors. This year, with Tenorio away, Vodafone’s key messaging fell to Head of Open RAN Francisco (Paco) Martin.
Martin said that Vodafone had rowed too far out into Open RAN to row back now. “The status is,” he said, “it’s ready for prime time. We are delivering not only from a KPI point of view, but also from a procurement perspective we can see one of the intitial goals, which was vendor ecosystem competitiveness, has been achieved.”
That would seem to give weight to the view that one of Open RAN’s chief benefits for operators was that it gave operators some element of competitive price leverage (and roadmap) over vendors in a world where they could no longer access Huawei.
Although it’s close to closing its tender, which made it hard to point to specifics, he remained committed to Vodafone’s original messaging. He also pointed out that the company continues to innovate in its Malaga centre as it looks for improvements in hardware technology.
Deutsche Telekom’s Petr Ledl said that the operator is still working on how to integrate multi-vendor solutions in the most effective way, to decrease the cost of deployment. He repeated the target to have 3,000 Open RAN sites in Germany by 2027. That has involved forming a “coalition of the willing” with AT&T and Vodafone to put together an advanced set of test plans that contains performance requirements – “the big deal about it is that we are aligned in agreement with other operators. So we can make sure that when solutions are tested it can be easier adopted by others who do not need to repeat tests again and again.”
In a panel hosted by TMN, Atoosa Hatefi said that Orange would proceed according to business drivers, having started with a phased approach with a RAN share and cloud RAN trials in rural areas. It plans to etend that to other areas in 2025, including in high capacity and urban areas, and then from 2026 it expects to see more volume deployments. So far Orange’s main focus has been on cloudification of the RAN platform, and vertical integration of Open RAN solutions. At the same time she wants to see more horizontal interoperability, particularly in Massive MIMO radios.