A couple of rabbits out the hat on the first day stage at FYUZ.
Firstly – does 2023 FYUZ further mark the rise of the incumbents in Open RAN?
So yes, we had Ericsson taking a priority slot in the keynote opening session to talk up its “Open RAN readiness” with VP Networks Fredrik Jedling coming in over video to step through its commitment from a couple of weeks ago.
Host Santiago Tenorio, not only Chair of TIP but also at the helm of Vodafone’s Open RAN efforts at group level, asked twice if Ericsson is truly ready for Open RAN. Jedling said indeed it has a million RUs deployed that are hardware ready. What this means is that these RUs have the silicon capability to do the Massive MIMO uplink processing that is required within the new category of fronthaul interface that Ericsson (and others) have pushed through O-RAN Alliance. This category puts more of the lower layer processing at the RU than previous versions, thereby taking away the requirements from the fronthaul. But to get to “Open RAN” operation, Ericsson needs to implement the Open Fronthaul category in software at its Cloud RAN sites and at the RUs. That is going to require a software upgrade that it will begin from 2024.
Ericsson is still playing canny about its ability to support third party RUs over that fronthaul connection. It has said it will do it on a customer requirement basis, but there are no commitments. So you have to ask, what is the point of Open FH if you are just linking Ericsson RUs to Ericsson Cloud RAN? Ericsson’s pitch is that by opening up horizontally it can enable greater automation in the network, exposing network capabilities by APIs, and get to those new 5G use cases and business models. Of course you can argue those benefits are more about Cloud RAN and cloud-native deployments and operations than they are about open interfaces between elements. Or you could argue that you won’t get the SMO and RAN automation beenfits without supporting O-RAN vertical interfaces between the RICs and the vRAN platforms.
Nokia has gone further – at least on third party integration. In a recent comment on LinkedIn responding to a TMN story, its head of networks Mark Atkinson said it had now proved interoperability with three RU vendors. A roadmap paper produced a couple of days ago repeats that claim, making the following points:
Nokia has already integrated third-party TDD RRH from three suppliers with its Baseband, and the solution is available today. Nokia AirScale Baseband is qualified for commercial Open RAN deployments for both TDD RRH (FR1) and mmWave (FR2).
• Nokia is currently integrating third-party FDD RRH with its Baseband; pre-commercial later this year and Commercial in 1H 2024.
• Once these initial integrations are complete, it will be possible to support commercial Open RAN deployments with third-party FR1 TDD RRH or FDD RRH across any frequency bands.
And, a year after announcing a partnership with Nokia on Open RAN, Vodafone had a new announcement that it would be deploying an Open RAN trial in Italy with Nokia. The deployment will see Nokia containerised baseband software running on Red Hat OpenShift for Kubernetes, on Dell PowerEdge XR8000 servers equipped with Nokia’s Smart Network Interface Card (NIC) for Layer 1 processing, which is powered by Marvell chipset. Nothing, thought, about third party integration horizontally, from DU-CU to RU.
Vodafone’s added that its deployment model, developed to support a 2,500 site re-do in the UK, now includes a third party Massive MIMO RU integration with NEC.
So we should add that Open RAN still includes challenger vendors – such as Mavenir, whose John Baker was at pains to point out on stage that in his reckoning there are 20 players in each of the main areas of Open RAN developement – RU, DU-CU and in the RIC, plus a good ten plus integrators.
The rise of the incumbent really depends on perspective. Are the incumbents bending to the will of the Open RAN operators, or is “Open RAN” bending to the incumbents?
Chips again – Intel be back
There’s another sort of incumbent, and that’s at the chip level. Intel, with its FPGA-based chip platform and lookaside acclerator made the early running, but it was a model that swiftly got labelled as too expensive and inefficient to achieve feature parity with integrated radio silicon from Nokia and Ericsson etc.
But Intel hit back in late 2022/early 2023 with its SapphireRapids release incorporating its vRAN Boost acceleration (more an of inline-like deployment, depending on who you believe). That got a big welcome a year ago from Rakuten, in the form of its now departed CTO and Symphony co-CEO Tareq Amin, who had been agitating for more openness between the lower layer vRAN software and the chip platform, as well as performance increases from the chip layer.
Today, Vodafone’s Tenorio said that Intel is coming back strongly into its R&D plans at its chip research facility in Malaga. A year ago, Tenorio was announcing co-developments with Marvel, Qualcomm and EdgeQ – and today’s Italy announcement referenced Nokia’s use of Marvell as its L1 chip provider. Tenorio wasn’t discounting Intel in 2022, far from it, but he was clear that the industry wanted to see gains in chip performance, and also more flexibility between that low PHY and the silicon.
But Intel is back in favour – giving Vodafone access to test sample silicon with integrated algorithms and applications, Tenorio said. The cost of access to silicon has been a blocker, yes. Smaller vendors don’t get access on the same terms as the big guys, and therefore their costs go up. A year ago Mavenir’s John Baker was one of those making that point. So perhaps we should see this partnership in those terms, Intel coming to the table with test silicon at a cost point that can enable partners to experiment.
There were more references for Intel. Dish’s Sidd Chenumolu VP, Technology Development at DISH Network said that he thinks the Xilinx platform that Dish started with was “more of a PoC”, rather than a full solution. He then praised Intel’s CascadeLake and SapphireRapids platforms as enabling a truly programmable and automatable platform. “I want the machine to do the automation, not me,” he said.
Manish Singh, Dell’s CTO, Telecom Systems Business, added his support, noting that Dell’s servers can now take advantage of Intel Sapphire Rapids for vRAN workloads in the right form factors and site based designs.
Integration
Last year at FYUZ we asked about integration: how much is too much, who does it, how open is open etc? If an issue can be an incumbent then it looks like integration as a hot topic is still very much around.
As noted above, Vodafone has a live deployable site integration, with Samsung vRAN software deployed in a Wind River cloud on Dell hardware, and now a third party RU. Andrea Dona, its UK tech lead, said that is a commercially-ready deployment that it can lift and shift into different territories. How swappable is it? We ask because Tenorio also told us that Vodafone is also in the midst of a global, 170,000 site tender, and remains commited to its intention for 30% of its site footprint to be Open RAN compliant. And, as we’ve seen, it is looking at Nokia’s deployment of its Marvell-powered SmartNIC integrated into the Dell server.
So… Vodafone has done the integration on its site design, but is also mid-tender for, what, 50-60,000 sites to go Open RAN. If Nokia pitches and wins, it is swappable into a Wind River cloud on a Dell server? If Ericsson pitches and wins (as it has done for the non Open RAN part of Vodafone’s Huawei site swap out in the UK) does Vodafone meet its self-imposed target by insisting on the open fronthaul interface as part of a Cloud RAN transition. If it does that, does it get to the benefits it says it is already getting from Open RAN in the UK.
These benefits are KPIs that show that its Open RAN sites are outperforming its legacy vendor. Dona claimed that already TCO is down 30% – “we need more” – and that it is already deploying a site in two and a half hours, instead of four. In time cloud-native automation pipelines might see that come down to ten minutes.
This hits on a key point about disaggregated network integration. Vodafone states it is looking for faster integration and testing compared to integrated solutions.
If the industry creates deployable blueprints, you can end up with a set integration chain that is easier for the operator to deploy. Rakuten Symphony and NTT DoCoMo with its OREX have spotted the opportunity to integrate solutions on a blueprint. But as you do this, you risk locking in integraton chains.
AT&T’s Rob Soni said that the operator previously found that vendors were coming to it with a preference for partners, this CaaS, this hardware platform, this cloud environment etc.
He wants to see the industry find a way to orchestrate the process of integration so things can be open and swappable, without each operator duplicating their integration efforts. And that works both ways, as Soni said. Vendors need to hear a consistent set of requirements from the operators, so they know what capabilities to build for.
A year ago, TIP announced that it was working on this very issue – creating certification and deployment badges and ribbons to create a marketplace of deployable solutions. An open question for this week – how is that process going and is it delivering?
Umit
Great insights for those who couldn’t be there, thanks Keith