Ericsson overhauls market approach with open RAN conversion

When Ericsson's top executives minus the supreme being of Börje Ekholm converged on London in early 2023 for their annual pre-Mobile World Congress event, the Swedish kit vendor was no friend to open radio access network (RAN) technology, a doctrine viewed outside Ericsson as a potential threat to its empire. Open RAN (or O-RAN) specifications would hurt performance, was the resounding message from Ericsson, which preferred to talk about its latest purpose-built kit.

What a difference a year makes. Fresh from landing a $14 billion contract with AT&T – not only the industry's biggest open RAN deal so far but also the largest contract in Ericsson's history – the same men and women arrived in London last week as converts, sounding like former sinners who had discovered the faith. Having seen the influence open RAN now holds over its biggest customers, Ericsson has decided resistance would be more damaging. As religious conversions go, this is pure realpolitik with ancient origins. Constantine saw how Christianity could support the Roman Empire's interests. Ericsson is similarly approaching open RAN.

In today's world of "closed" RAN, suppliers largely resemble neighboring geographical fiefdoms within a telco network. In each of these fiefdoms, one company has absolute control at all levels. Any supplier diversity, for a telco, comes from ensuring there is not one fiefdom across the whole network. But at any given site, a single player is all-powerful.

Open RAN was conceived partly as a fix. By scrapping the proprietary interfaces between different parts of the RAN, and replacing them with an industry standard, operators would be able to buy parts and software for a site from multiple suppliers instead of one vendor's system. That would make them less beholden to big vendors like Ericsson. Yet even the telcos most enthusiastic about open RAN have acknowledged the economic and operational drawbacks of hosting a banquet of suppliers rather than a small private gathering.

Under a new paradigm, proposed by Ericsson in London, the geographic fiefdoms are swept aside. Instead, the network looks more like a system of government where some responsibilities are under central authority, others are devolved and all can be voted out. To use a food analogy instead, and one Ericsson might appreciate, the network is a Scandi-style open sandwich with a single layer of bread beneath an assortment of other comestibles.

"I think the industry has been characterized by vertically integrated geographical splits and we believe that paradigm is gone with this shift," Fredrik Jejdling, the head of Ericsson's networks business, told Light Reading. "Therefore, the competitive premise shifts from being integrated and vertical to being strong and horizontal in the layers."

A smorgasbord of radios

How this could look in practise is illustrated by the AT&T deal. Until now, the US operator has allowed fiefdoms to exist, deploying Ericsson across two thirds of its RAN and Nokia for the rest. But Nokia is being evicted. And while Ericsson is not expected to be a 100% replacement, it is down to provide the foundational layers across the whole of the AT&T footprint – the bread slice, if you like, in the open sandwich. "AT&T sees benefits in operational efficiency of going to a single vendor," said Erik Ekudden, Ericsson's chief technology officer, during an interview in London.

This horizontal layering probably means that AT&T will forever have only one set of baseband software across the whole network. Besides connecting some of its own radios to this software, as it would in a fully integrated RAN, Ericsson will also link to Fujitsu, the Japanese vendor (already an Ericsson partner) cited in the AT&T release, and possibly others in future. Service and management orchestration (SMO) seems likely to come from Ericsson, too. "Ultimately, those types of horizontal networks start with the RAN software, with the management and orchestration," said Jejdling.

Ericsson's earlier concern about open RAN specs stemmed from the initial decision by the O-RAN Alliance, the group behind the concept, to put most functionality in the distributed unit (DU), the server box for baseband processing, instead of the radio. Ericsson's experts reckoned this would lead to performance problems with antenna-rich massive MIMO technology. Under compromises reached last year in Japan, vendors can move important uplink functions to radios and still be deemed compliant with specs. But their baseband units must include support for those functions regardless.

While Ericsson has ruled out developing more basic radios, AT&T should – in theory – be able to procure those from another vendor and connect them to Ericsson's baseband. "Our baseband supports having equalization algorithms in the baseband as well as in the radio," said Jejdling. "It is in our roadmap to support both configurations."

The paradigm of having one software layer to unite radios from multiple vendors would appear to suit other telcos, as well. Apparently frustrated by the multitude of radio vendors already in his footprint, Iain Milligan, the chief networks officer of the UK's Three, expressed interest at a recent press conference in Glasgow in using DU software from Mavenir, a US vendor, across radios from Ericsson, Huawei, Nokia and Samsung. For Yago Tenorio, Vodafone's head of network strategy, open interfaces are needed mainly to boost competition in the radio subsector.

"It is very important that we get Ericsson and Nokia to do it because if you open interfaces then apart from having more software vendors you can have ten times more radio vendors," he said during an interview last year. "That will not be limited to just Nokia, Ericsson and Samsung. You will have Mavenir, Fujitsu, NEC, even ZTE where you can use them. That is where the majority of the money is going and that is a layer where you want to have competition."

One platform to rule them

The obvious criticism of this new Ericsson paradigm is that it could leave an operator even more dependent on a single vendor if it is the sole supplier in a particular horizontal layer. As far as Ericsson is concerned, the openness of the latest specs is the safeguard against this kind of tyranny. "You have the possibility to change Ericsson to someone else," said Ekudden with specific reference to AT&T. "There shouldn't be any reason for them to make that choice because then we are not doing our job in terms of always investing enough to lead in the market."

A switch, nevertheless, could be tricky for commercial and operational reasons. Integration of baseband and radio technologies is expensive and difficult to do well, according to numerous experts, even with open interfaces. This is largely what Ericsson alludes to when it talks about "industrializing" open RAN. "It is honestly not that difficult to integrate one radio into one baseband," said Jejdling. "Doing it with cost structure and performance structure is a whole different ballgame."

Introducing additional vendors in those layers to avoid overreliance would upset the paradigm, and it would probably raise costs (and complexity) by increasing the number of suppliers across the entire network. "Having more vendors and having too much fragmentation will not bring any kind of advantage cost-wise or operationally," said Ekudden. "You don't want to have too many."

Where openness ends

Another concern is that standards and interoperability seem less advanced in some areas than others. Ekudden breaks the open RAN interface discussion down into three categories – the cloud (where the interface is between hardware and software); SMO (artificial intelligence and the RAN); and fronthaul (baseband and radios). Ericsson and others seem to have moved past concern about fronthaul. But the same cannot be said about cloud and SMO.

Japan's NTT Docomo and Vodafone were arguing in a white paper published as recently as October that openness ends with SMO, for instance. As the label implies, SMO is supposed to aid more automated maintenance of network functions, cloud infrastructure and the RAN intelligent controller, a kind of app store for the network. It should, ideally, replace the vendor-specific element management system (EMS) used in a closed RAN. Yet critical interfaces such as O1, O2 and A1 were still half-baked, according to NTT Docomo and Vodafone in their October white paper.

As for the cloud, the big vendors are still divided along religious lines about the use of general-purpose processors. Nokia, essentially, deems these insufficient for the resource-hungry "Layer 1" RAN software, which it offloads to custom Marvell chips. Ericsson disagrees and says – perhaps not unreasonably – that its approach is more hardware-agnostic. "Ericsson software is truly independent of the underlying hardware," said Jejdling, acknowledging the alternative technique has its attractions. "That is how we want to design it – so that it provides a level of openness that we believe our customers are asking for. "

Yet another cloud issue is the ability to host software on any cloud platform, whether a hyperscaler stack or one built by containers-as-a-service (CaaS) specialists such as IBM-owned Red Hat and Broadcom-owned VMware. Nokia has given up work on its own infrastructure platforms, saying this was designed to host only its own applications when operators want to have single platforms for all third parties and avoid cloud silos. But Ericsson has yet to abandon its own cloud-native infrastructure solution (CNIS). Asked where Ericsson's RAN software would sit, AT&T said functions would be divided between Microsoft Azure and CNIS.

Still, the survival of CNIS in its current shape must be in doubt. BT, notably, has deployed Ericsson's 5G core software on its own telco cloud, cutting CNIS entirely out of the picture. "We need to talk to Ericsson about what their plans are," said Howard Watson, BT's chief security and networks officer, at a recent press update. "They would automatically focus on their own in-house solution first and then look at other industry solutions. Getting beyond that was initially hard, but we've achieved it."

He would be encouraged by the words Jejdling spoke just a couple of weeks later. "Our ability to build a flexible software structure on top, irrespective of the underlying CaaS or RAN compute, is going to be important," he said. "We need that flexibility in the way we do software in the future."