'Open RAN' now justifies using a single vendor

Open RAN is the emperor's new clothes of telecom. In this modern-day twist on the famous Danish folktale, AT&T is the emperor of the story claiming to be dressed in the finest open RAN suit that money can buy. Everyone can see AT&T wears nothing of the sort and is nakedly exposed to Ericsson, but hardly anyone says it, fearful of angering the emperor or appearing stupid. Instead, there is much toadying. Even people who previously attacked Ericsson, the tailor of the story, as an open RAN charlatan have joined in the praise.

The entire concept has now been turned awkwardly on its head. At first, open RAN was about combining multiple vendors at the same mobile site to boost competition and cut dependency on a single supplier. Now it's presented as an insurance policy against lock-in, a get-out-of-jail-free card to justify a single RAN deal. The logic sounds brilliant. If something goes wrong or Ericsson starts to fall behind its rivals, AT&T can just swap out the problematic parts. It wouldn't have been able to do that before the Swedish vendor's miraculous Damascene conversion to open RAN.

Saying is one thing. Doing is another. For a start, 5G equipment, especially the advanced massive MIMO sort, is notoriously expensive. No telco pleading poverty, as many do these days, is likely to pay for a swap before it has reached the end of its useful life. If there were a malfunction, the provider would be expected to fix it. If there were a Huawei-like ban, the whole system would have to go. Open RAN is not currently a plug-and-play technology, and it probably never will be. Systems integration is time-consuming and hard. Unless multivendor is built in from the outset, it probably won't happen.

Cloud cuckoo land

AT&T, however, identified no other vendors that would allow this to be described as a multivendor open RAN deal. Dell belongs in the cloud RAN camp, as does Intel, which might also be named because it is down to manufacture custom chips for Ericsson in future. Both AT&T and Ericsson have also confirmed that custom chips will be used. The initial rollout, apparently, will have no need for cloud RAN kit.

"To start, we will leverage the strength of Ericsson's silicon roadmap to introduce open radios onto our network," said AT&T via email in response to questions. "As silicon solutions mature, we will introduce additional open architecture solutions based on the Intel roadmap." But deploying purpose-built kit will hinder the introduction of cloud RAN at those sites – unless AT&T plans yet another costly swap-out before equipment is due to be replaced.

Why bother virtualizing or cloudifying anyway? Experts say the big attraction is being able to share IT resources, which partly means running all the workloads on the same cloud platform. But AT&T won't even be doing this. "We will use Microsoft Azure for some functions and Ericsson's CNIS [cloud-native infrastructure solution] for other functions," it said. This raises questions about the ease of putting Ericsson's tech on Azure or third-party apps on CNIS. Regardless, it will leave AT&T looking slightly siloed. And two platforms will be more expensive to maintain than one.

As for other named suppliers, Corning and Fujitsu were already reckoned to provide distributed antenna system technology to AT&T. Fujitsu's radios could always be synchronized with Ericsson's baseband. But Ericsson and Fujitsu have been 5G partners since 2018, the year the O-RAN Alliance was founded, if not before. And AT&T clearly intends to use a lot of Ericsson radios, its public discourse shows. It declined to say how many sites are to feature the Japanese alternative and evidently has no firm plans to use anyone else, saying there could "potentially" be others in future.

The single RAN nature of this deployment means the rest of the industry must simply trust AT&T and Ericsson when they say their network will comply with open RAN specs. And unlike Nokia, which recently hailed compatibility with Mavenir and three other vendors, Ericsson has provided zero public evidence of interoperability tests (Light Reading went exhaustively through Ericsson's press releases so you don't have to).

Nor does the O-RAN Alliance, the specifications group led by AT&T and four other telcos, have anything about Ericsson interoperability tests on record, it confirmed to Light Reading. At Open RAN North America – an event organized by Informa (Light Reading's parent) and held last week in Texas – Mavenir, Nokia and Samsung all indicated they have never had the opportunity to test open RAN interoperability with Ericsson in any forum.

Massive MIMO mumbo jumbo

So exactly what interfaces will be used to connect Ericsson's baseband to Fujitsu's radios? Asked that question, AT&T said Ericsson will support both 7.2 "Cat A," the original open fronthaul spec for basic radios (that is, ones featuring up to eight transmitters and receivers), as well as the new "ULPI" modification for more advanced "massive MIMO" gear. But this is where it gets complicated.

ULPI stands for uplink performance improvement and came about when Ericsson and others started to complain that 7.2 "Cat B," the spec originally conceived for massive MIMO, wasn't up to the job. Too much uplink functionality had been pushed from the radio into the distributed unit (DU), a server box for baseband processing, they said. Performance could suffer as a result.

Two proposed fixes emerged, each with big sponsors. And because no one was prepared to back down, the O-RAN Alliance effectively adopted both in an awkward compromise – what some might uncharitably call a fudge – earlier this year. Class or "operation mode" A, Ericsson's preference, shunts the equalizer and other uplink functions into the radio. Class B transfers some uplink functions to the radio but keeps the equalizer in the DU.

All DUs, however, must include equalizers to ensure there is compatibility with the different flavors of open RAN. It is hard to see how this duplication between DUs and class A radios will produce lower costs, one of open RAN's promises, and Ericsson has confirmed it will not be manufacturing radios based on either class B or the original Cat B.

None of this, incidentally, means class B is necessarily better. Rather than duplicating the equalizer, it duplicates channel estimation to avoid sending data from the radio to the DU, a transfer that could overload the fronthaul link. But if the DU vendor differs from the radio vendor and their algorithms don't match, there might be a problem.

This leaves the industry with a mishmash of "standards" for open fronthaul – Cat A, Cat B, class A, class B. It is far from ideal. "The challenge for DU vendors is to support all four interfaces on the DU, which Mavenir is committed to do," said John Baker, Mavenir's senior vice president of business development, in a LinkedIn post. "If the operator chooses a DU supplier that does not support all interfaces, they will be locked to that supplier."

Oh, the irony

There has long been suspicion within some quarters that AT&T and its big telco siblings are not super-serious about open RAN new entrants, and merely using the threat of an open RAN swap to extract lower prices from traditional vendors. The deal with Ericsson has clearly been struck on very generous terms, according to Earl Lum, an analyst with EJL Wireless Research. He even reckons it includes "vendor financing," which would effectively mean Ericsson bought the contract.  AT&T has, however, dismissed speculation about staff joining Ericsson. "No employees are moving as part of this agreement," it said.

Yet for companies perceived to be overcharging their customers, Ericsson and Nokia are hardly in the best shape. Ericsson recorded a net loss of 30.5 billion Swedish kronor (US$2.9 billion) for the recent third quarter and a margin of just 3.8% for earnings (before interest, tax and amortization). Profits at Nokia fell €295 million ($318 million) year-over-year, to €133 million ($143 million), for the same period. That was before the AT&T decision, which has further weakened it. Between them, the Nordic vendors have this year announced plans to cut up to 22,500 jobs, about 12% of the total.

The irony of the deal, then, is that it could act as a driver of consolidation, leaving the industry with even fewer suppliers. If other big telcos apply the same logic as AT&T, Ericsson or Nokia – depending on which has the bigger footprint at the operator in question – will take over whole networks, promise to be on their best open RAN behavior and show up with small knapsacks of third-party radios to demonstrate compliance. After last week's developments, any small vendor already struggling to make an impact should be alarmed.