As large operators in western Europe and Japan become increasingly enthusiastic about open network platforms such as Open RAN, it sometimes seems as though the US majors, which were early drivers of the disaggregated open platform, have become less supportive.
However, judging by recent comments from AT&T and Verizon, that is not the case. Both are making significant progress in deploying RAN, core and transport networks based on white box hardware, edge computing and cloud-native architectures, and both expect this to start having a visible impact on their cost base and service agility from 2022 or 2023.
But where they do diverge from the European giants like Vodafone and Orange is in having a lower focus on multivendor networks that can be deployed easily and automatically. That may be the eventual goal, but in a head-to-head that has always centered heavily on network performance differentiation, AT&T and Verizon are working more like advanced Asian operators than cheerleaders for a common, plug-and-play platform.
Both vendors have invested heavily in developing their own specifications and frameworks for virtualized and software-defined networks. AT&T, in particular, has contributed a lot of its work to open initiatives such as the O-RAN Alliance and Open Compute Project, but it is not deploying its own networks based on vanilla specifications from such bodies, but using systems developed by large inhouse teams in conjunction with vendor partners.
This is how the two US giants look like Asian leaders like NTT Docomo or China Mobile, which have, for several mobile generations, sought performance differentiation by working closely with a tight-knit circle of favored suppliers to create highly customized platforms that might, or might not, then be shared with others via standards bodies or open initiatives. Verizon has always worked this way too, most recently in devising its own pre-standard specs for its first 5G network, and before that for its IMS (IP Multimedia Subsystem).
And it is notable that, for now at least, the inner circle of vendors looks very traditional, with a few shining exceptions, such as AT&T’s relationship with Drivenets in the disaggregated, white box transport network. AT&T has certainly tried to shake up its supply chain over the past years, since launching its SDN 2.0 programme, which introduced some new names, such as tail-f. However, tail-f was quickly acquired by Cisco, reintroducing the router leader into the AT&T fold – perhaps a source of hope or fear for the current wave of companies targeting a place in the 5G core or RAN.
Verizon’s networks SVP Adam Koeppe said the operator will start taking delivery of Open RAN-compliant equipment later this year, and that most of the shipments by year end will support at least those specs that are commercially stable (likely to be led by the O-RAN Alliance open fronthaul standard).
However, it does not seem that this open equipment will open up Verizon’s existing 5G supply chain. Koeppe said the compliant kit would come from Ericsson, Nokia and Samsung, which begs the question of how far Verizon will be going towards the fully-fledged Open RAN architecture, since Ericsson supports very few of those specs commercially.
But in future, the breaking of the fronthaul lock-in between radio and baseband unit will give Verizon the option to introduce third party radios to its existing 5G network, and eventually to deploy RAN software from new suppliers on its radios and basebands.
“If you’ve got a new company that’s providing phenomenal radio equipment, that’s different from what’s in the market today, when they’re all making O-RAN-compliant equipment and following the specifications of O-RAN, you can choose to bring in some of that best-in-class equipment from maybe a non-traditional partner,” he told Light Reading. “So that’s really the intent behind having that openness on the radio access network.”
AT&T’s latest move towards an open, virtualized RAN – which it has said is a goal for 2025 – came with a vRAN test in New Jersey. Again, this was supported by a traditional supplier, Nokia, which provided its new vRAN 2.0 system, complete with O-RAN specs.
The two companies ran a proof of concept test, with an end-to-end Layer 3 data call using a fully virtualized distributed unit (vDU) and a centralized unit (vCU), supporting different applications including web browsing and video playback.
Like Verizon, AT&T said the trial might feature a traditional vendor, but it was a step towards open RAN because “it facilitates decoupling of baseband hardware and software and allows AT&T to tap into an open, multivendor commodity hardware ecosystem for baseband compute and acceleration”, as Mazin Gilbert, VP of advanced technology, put it.
Nokia provided its AirScale vRAN 2.0 baseband with vDU and vCU running in the cloud, plus a fronthaul gateway and packet switch that were converted to connect existing CPRI radios to Ethernet-based enhanced CPRI (eCPRI), supporting the Split 7.2 fronthaul specified by the O-RAN Alliance.
“AT&T is working towards more extensively testing the performance, power, and operational considerations of vRAN in a limited production setting on live traffic to develop a data-driven understanding of the impact of vRAN for a brownfield operator such as AT&T,” Gilbert said.
Nokia and AT&T are also co-developing their own implementation of a key O-RAN Alliance technology, the near-real time RAN Intelligent Controller (RIC).
Gilbert told SDxCentral that vRAN would be a “generational opportunity to refactor legacy RAN software to a modern, cloud-native design” and will be particularly important to support AI/ML for RAN control and optimization. That, in turn, will enable operators to “use their wealth of RAN operational data and customer insights to customize their networks using data science and AI technologies to their particular customer base, geography, and spectrum position.”
He acknowledged that “there is a bit of work to do to catch up in feature parity with the legacy stacks that have undergone decades of development”, but was optimistic that processing power was developing rapidly to enable vRAN to “approach and surpass custom-built solutions … AT&T is closely monitoring and participating in these ecosystem developments and will make appropriate decisions to scale vRAN more broadly on its commercial 5G network as the technology matures.”
Meanwhile, AT&T’s CTO Andre Fuetsch was certainly not prepared to pass the open networks baton to other telcos, and hit back against claims by some ecosystem players that AT&T had gone cold on fully open architectures. He told a recent conference that AT&T had invested $135bn over the past five years in its US networks and that was with an automated, self-healing architecture and open specs in mind.
“A lot of this was started with a program about seven years ago when we set out to virtualize and software-enable the majority of our network. And this is really the foundation of everything we do, and part of this really important foundation is open source technology,” he said. “The advances we’ve made in cloud-native virtualization, containerization, hardware and software disaggregation have really allowed us to reach our software based networking goals of 75% … right now we’re well over 77%.”
Fuetsch highlighted six open source projects that he said were foundational to AT&T’s network vision for the whole of the 2020s. These are:
The Facebook-founded Open Compute Project (OCP), to which AT&T has contributed designs for white box routers and cell site gateways.
Airship and the Cloud-Native Computing Foundation (CNCF). Airship supports a network cloud infrastructure that integrates 14 CNCF projects in order to create AT&T’s Network Cloud.
Anuket, which merges the Common NFVi Telco Taskforce (CNTT) and the Open Platform for NFV, to provide reference cloud infrastructure models and architectures.
Open Network Automation Platform (ONAP), a Linux Foundation project that was originally based on AT&T seed code for management and orchestration of virtualized networks.
The O-RAN Alliance, also partly founded on AT&T seed code from its xRAN initiative. Fuetsch said the proof of concept test with Nokia was a “key milestone” in proving the maturity and capability of open architectures.
“The goal of open RAN is to really enable new participants to come into the RAN space by disaggregating the RAN with open interfaces, and the open RAN software community continues to grow it,” Fuetsch said.