One of the live questions this week at TIP’s FYUZ event has been the extent of Ericsson’s relatively newly-stated commitment to Open RAN.
The vendor announced that it will start rolling out software to support the latest enhanced O-RAN fronthaul interface between its Cloud RAN software (vDU-vCU) and Massive-MIMO RUs.
But questions were asked if it would go further and also support existing O-RAN fronthaul interfaces, so that non M-MIMO RUs, and other vendors’ M-MIMO RUs, could also attach to an Ericsson Cloud RAN baseband (also known as the vDU-vCU).
It says yes and no. Ericsson has previously told TMN that it would consider connecting its DU to third party RUs on a customer need basis. Now it has confirmed that it will support Cat A, which is the Open Fronthaul interface that links RUs with less than 8T8R MIMO. It will also, as announced, support its own proposal for the new enhanced interface from the DU-CU to Massive MIMO RUs – known as Uplink Performance Improvement (UPLI).
But Ericsson has also confirmed to TMN that it will not be supporting the existing (un-enhanced) Cat B specification for O-RAN Fronthaul as it does not think it provides adequate performance.
It sounds esoteric. Does it really matter which variants of O-RAN fronthaul interface a DU supports? Some say it does, and that Ericsson is attempting to stall the market for rival RU vendors. What Ericsson is saying is that if RU manufacturers want to be deployed in a network where Ericsson is supplying the vRAN, they will need to design their M-MIMO RUs to the new specification. Of course, with a million of its own RUs already “hardware ready” (ie. they have the on-board silicon) for the category, Ericsson can leverage its installed advantage to its benefit.
Santiago Tenorio, Network Architecture Director, Vodafone, said that in his opinion Ericsson “has a point” on the need for a new split, which splits uplink processing (call it an equaliser for shorthand) across the RU and DU. But he added that he still wants to be able to support all possible combinations of RU-DU.
“I think ending up with the equaliser in two places is a necessary compromise. I tend to agree with their position that is it better to have it on the RU than on the DU because it helps interoperability. If the equaliser is on the RU it gives you easier interoperability with the DU than if the equaliser is coming from a different function.”
Tenorio added, “Yes you can try and drive everyone to the same class category, but the most important thing is to make sure that the ecosystem works in all possible combinations – so the DU is ready to take on whatever category, whatever class.”
So even though Tenorio agrees with Ericsson, he would still look for support for other options?
“Yes because the situation is what it is today. You need to make sure all possible combinations work together and you’re not going to run into camps. That would be bad. What you have now is different choices, but the standard is making them all compatible. In simple terms imagine the RU has the equaliser, in that case you can skip the equaliser in the DU and then that’s it. So you can make two different classes work together by bypassing the functions that overlap. And it’s just an equaliser, it’s not a huge duplication of cost.”
Maite Aparicio, Head of Open RAN, Telefonica, which recently announced it would be working with Ericsson as a Cloud RAN partner, was another operator to offer support for Ericsson’s fronthaul choice.
“We are finally able to find a solution that is there. Many voices are against ULPI but finally we are able to find a solution and we need to be positive in this as we finally find a solution for a problem we were discussing for months.”
John Baker, SVP Business Development, Mavenir, speaking at a panel at FYUZ, said of the new interface, “At the end of the day it’s an edge case for Massive MIMO. Cat A and B are perfectly OK for normal radios and for Massive MIMO. So then do I put more cost in the radio or do I keep it in the baseband? In reality you want it in the baseband not in a radio at the top of a tower. I can see that they [Ericsson] can use it to delay things, but why would an operator want silicon in the tower for a fraction of -dB edge case performance.”
In further comments to TMN, Baker pointed to quotes from Tenorio in a LightReading interview which stated that a Mavenir RU was showing “dramatically better” energy efficiency than anything Tenorio has seen from the big incumbent legacy RUs.
If Ericsson does not support the eixsting Cat B from its DU, then operators deploying its Cloud RAN would be cutting themselves off from the “dramatically better” RU, Baker said.
“The Mavenir radio today is CAT B, as the Ericsson specification has not been implemented in our radio yet. Why would you cut off the opportunity to use a better radio just because Ericsson does not want to implement the interface? Pick another DU which could include Mavenir and problem solved.”
The other option, of course, is for Mavenir to support ULPI from its own RUs, which it has actually committed to doing. Indeed Baker says, “We support the first two categories already, and when this one comes out we’ll support it.” But going by his earlier comments, it’s not a deployment option that he’s thrilled about, seeing it as an edge case offering only limited performance improvement, whilst incurring more cost.
Open RAN in name only?
Tenorio rejects the cynic’s view that by adopting this new interface, Ericsson networks can be badged Open RAN in name only and thereby solve the operator’s self-imposed Open RAN targets.
“What do we gain by that? We never started Open RAN for that. We started for optionality in the supply chain, so we need to get that. Yes it’s good news that Ericsson is going Open RAN and we hope we can drive that to be as good as the one we are rolling out in the UK. But we need at least one if not two other alternatives ready.”
Tenorio adds that he is not taking a cycnical view of Ericsson’s commitment.
“I think Ericsson has plans and they’re firm. Some in the industry are still sceptical and try to read between the lines of what they say.
Time will tell. We will take it through the same process as we took the other ones. We’ll take it to the field, try solutions, look at the roadmap and try to improve that. And we can use our RFQ to try to crystallise commitments and improvements.”
Tender is the might
Ah yes. The RFQ. Tenorio this week backed up a four year promise by stating that its global RAN site tender, due as its current contracts end in March 2025, will see 30% of its 170,000 global sites be Open RAN compliant.
How is that going to work?
“We need a new contract for every market and every site. Even if the market stayed the same on single RAN at every site, we still need new contracts to keep on buying kit. So that’s a tender. And then once we have a tender we need to think of a strategy. Ours is still to introduce 30% of Open RAN. That would be a reasonable outcome: in some cases that will mean shuffling the allocation of the market to different players and in some cases it might stay the same.
“It’s obvious that we would make sure that every player has an incentive and a threat. It’s a natural thing to do – to make sure that every player has something to win and something to lose. 170k sites are up in the air. It’s a big bet and you need to drive competition. If you start with the allocations that vendors keep what they’ve got, you won’t get the best prices.”
