Open RAN and its multivendor misfortunes

Open RAN started out as a noble attempt to pair radio units (RUs) from one vendor with compute products from another. Since then, it has spawned enough letter-and-number combinations to resemble one of the Enigma key sheets issued to Alan Turing's code crackers. These A1, E1, E2, F1, O1, O2, X2, etc., etc. interfaces cover just about every imaginable bit of joinery holding together the parts of a radio access network. But the effort increasingly seems to outweigh the benefits.

Take that open fronthaul 7.2x interface between RUs and the distributed units (DUs) hosting RAN software. It is replacing an older interface called CPRI, maligned because it meant any telco had to buy these components in a pre-integrated package from one vendor's system. But 7.2x doesn't eliminate the expense and hassle of systems integration. Parts cannot simply be slotted together like Lego blocks.

What's more, concern about the suitability of the original interface for massive MIMO – an antenna-rich 5G technology – has now birthed several mutations: 7.2x category A, for basic radios; 7.2x category B, for massive MIMO; 7.2x class A (one option to address the perceived shortcomings of category B); 7.2x class B (another option aimed at the same problem). If casual readers are confused, industry people in the know sound unhappy.

"We worry a great deal about fragmentation of the ecosystem," said Alok Shah, the head of strategy business development and marketing for the networks business of Samsung Electronics America, when he met Light Reading a few weeks ago at Mobile World Congress (MWC). That class A option involves putting more intelligence into radios and only so many vendors have the means to do it, he said. (Ericsson is regarded as the main sponsor of this approach.)

Massive MIMO mumbo jumbo

Ericsson gave its perspective on the open RAN dilemma in a recently published handbook on massive MIMO. "Open interfaces does not equal multivendor," it said. Besides ruminating on the costliness of systems integration, and the time it takes, the Swedish company pointed out that vendors collaborating on massive MIMO would need to coordinate software releases, including all necessary testing and verification, to guarantee interoperability.

In addition, the features developed would be dictated by a "minimum common denominator" in a tie-up of this nature, resulting in "performance limitations," according to Ericsson. An operator would also face the problem of identifying which vendor is responsible for any faults that arise (and therefore any fixes that need to be made).

Nokia is seemingly in agreement about the technical challenge. Complex algorithms split between DUs and RUs would have to match, forcing one or both vendors to make software changes, said Tommi Uitto, the head of Nokia's mobile networks business group, at MWC. Any vendor adapting algorithms like this could end up with multiple software tracks.

Today, massive MIMO is strictly single vendor. At least, there is no evidence of a significant "multivendor" deployment based on open RAN specs. What seems remote is the possibility of such partnerships between the dominant RAN vendors. Ericsson and Nokia seem as predisposed to coordinate their software releases as Donald Trump is to share details of his election campaign with Joe Biden's team. Nor is it obvious why operators or regulators would welcome such coordination by companies that collectively account for most of the RAN market outside China.

The problem of scale

Is this a problem? If open RAN's only achievement was to facilitate a coupling between Ericsson and Nokia (or other Tier 1 vendors), it would fail to realize its main objective of spurring competition and diversity in the market. Tie-ups have, conversely, tended to involve either a big player and a smaller one or two smaller companies. But some of these arrangements are starting to look as formal as joint ventures, the fruit of long sessions to produce a blueprint or template that can be taken around the world. They are as "pre-integrated" as anything out of Ericsson or Nokia.

Small companies have also failed to land big deals outside a few ailing greenfield projects, and the RAN market has been shrinking. Total sales dropped 11% last year and they are expected to fall by 4% to 6% in 2024, according to market-research firm (and Light Reading sister company) Omdia. Even if the market recovers next year, nobody expects anything but modest growth in the foreseeable future. Is the sector once again ripe for consolidation?

"It is a scale business," said Fredrik Jejdling, the head of Ericsson's mobile networks business group, when recently asked that question. "It takes a relatively large R&D investment, and it takes an investment in manufacturing and in supply-chain management to be able to deliver this critical network that 5G provides for us."

Consolidation is clearly desirable in Europe's fragmented service provider market, according to Ericsson, and Jejdling draws a parallel between that and his own sector. "We have a handful of players left," he said. "Fifteen or 20 years back there used to be multiple players in our segment as well. It is the same with semiconductors and the same with the cloud, and if you artificially try to keep up the number of players, each one is going to be subscale and then no one is going to be able to reinvest in the networks that are critical for the future of 5G."

Nowhere in his remarks does Jejdling call for another round of mergers involving new RAN product suppliers. But his comments do beg questions about the durability of "subscale" RAN vendors. An estimate from an AT&T technology executive last year put Ericsson and Nokia's combined annual R&D spending in this area at about $5 billion (together they spent $9.3 billion on total R&D last year). To look competitive, then, a company would have to be investing about $2.5 billion a year.

This looks impossible for all but a handful of players that belong to large Asian conglomerates – namely Samsung, NEC and Fujitsu. Of course, part of the rationale for open RAN was that a company with a much smaller budget could be competitive if it specialized. But the specialists seem to be dying out. Mavenir and Parallel Wireless, for instance, started out as software developers but subsequently expanded into hardware design.

Radio gaga

Diversity, nevertheless, looks more feasible and perhaps more desirable on the radio hardware side. NEC and Fujitsu are both recognized for their RUs but have struggled to interest telcos in their software for DUs or central units (CUs), where other RAN functions are handled. Mavenir has been able to land RU deals since entering this part of the market. Genuine specialists such as Taiwan's MTI remain in contention.

There may be a few reasons. To start, RUs attract more than 50% of the "wallet share," according to Yago Tenorio, the network architecture director of Vodafone Group. That would explain why telcos want optionality and means there might be room for a bigger number of players in this subsector.

Second, the supplier of DU/CU software often takes the lead in systems integration, according to Nokia's Uitto, and systems integration demands considerable resources, making it harder to compete for this business. Third, and perhaps influenced by these factors, open RAN operators tend to be combining one DU/CU software vendor with multiple RU suppliers and not the other way round.

No doubt, this model has been aided by open RAN specs that shift intelligence into DUs and keep RUs as basic as possible. Provided the DU/CU software vendor is good enough, operators can buy RUs from multiple suppliers to address various niche requirements. But as a growing 5G usage drives telcos toward massive MIMO products, there is a danger of complexity. If big operators ultimately prefer the class A split, smaller vendors could lose out.

Orchestration maneuvers in the dark

Meanwhile, many of those other Enigma-like interfaces do not look very open. Of particular concern is a layer called service and management orchestration (SMO) and the joinery that surrounds it. SMO is deemed a big differentiator by open RAN enthusiasts. It is supposedly the automated conductor to the many instruments of the RAN and comes with features like the RAN intelligent controller (RIC), a kind of app store for the network.

As recently as late 2023, however, Vodafone and NTT Docomo were warning about half-baked interfaces between SMO and other parts of the RAN. Without progress, "vendor proprietary extensions will be required to interconnect vendor functions with SMO, excluding SMO from the positive evolution," they said in a detailed white paper.

Ideally, an operator should be able to use one vendor's SMO in unison with another supplier's network functions. Ericsson says this is possible with its own SMO, branded the Intelligent Automation Platform (IAP). But there is still no evidence it has been done. And the $14 billion contract it announced with AT&T last December, Ericsson's largest "open RAN" deal so far, covers both SMO and DU/CU software.

When a vendor provides both, it has an opportunity to create "sticky functionality" between them, says Nokia's Uitto. If he is right, then switching from Ericsson to another supplier in future could be even harder. Moreover, AT&T's contract also has Ericsson down to provide its cloud-native infrastructure solution (CNIS) as the cloud platform for the RAN. The O2 interface between SMO and this layer was described as "partially vendor proprietary" in the white paper from Vodafone and NTT Docomo.

AT&T's use of IAP might also rule out the deployment of certain RIC apps. Until now, these have come in two flavors: rApps, for non-real-time functionality; and xApps, for the near-real-time sort. But IAP is good only for rApps. That is possibly because Ericsson sees third-party xApps as a threat to its mainstream business, said James Crawshaw, a principal analyst with Omdia, when IAP was launched back in late 2021.

One of the most prominent xApp developers is a company called Cohere Technologies. Ray Dolan, its CEO, made his name at Flarion Technologies before selling the highly regarded radiotechnology player to Qualcomm for $600 million in 2006. At Cohere, Dolan has pioneered a "beamforming" system that promises to boost the capacity of deployed radios. Operators including Vodafone have been impressed after field trials. But supercharging basic radios could endanger sales of the massive MIMO radios Ericsson now spends billions to develop. Today, you won't find Cohere anywhere near IAP.

A new category is now emerging, said Alex Choi, a senior Deutsche Telekom technology executive who currently chairs the O-RAN Alliance, the industry body responsible for specs. The dApp, or distributed app, would run directly on CUs and DUs, he said in a LinkedIn post, and satisfy needs that rApps and xApps cannot. "They handle real-time control loops below 10 milliseconds, focusing on radio management, resource scheduling, beamforming, and modulation functionalities," Choi said. Get ready for more interface mayhem.