Special Report: Open platforms in 5G
We have commented before on how far the move towards open platforms within mobile networks will also introduce open source models to the previously closed world of cellular infrastructure. To some extent, this debate is skewed by the market focus on the open RAN – understandable, given that the RAN is by far the largest element of mobile expenditure, but not always representative of developments in other areas of investment.
So far, open source models have made limited impact on the RAN. The O-RAN Alliance does have an open source strand, called O-RAN Software Community, which has some projects that are additional to the main program (including work on Split 6, which has been taken up on a more commercial basis by Small Cell Forum). But the decisions made by the Alliance’s leading contributors tend to inform its priorities, hence the focus on Split 7.x for Open Fronthaul specs, and open source has a very limited role.
However, this can obscure the fact of the rising impact of open source, particularly projects hosted by Linux Foundation, on the broader mobile telecoms landscape. Many of the core technologies that are propelling new open platforms in transport networks, software-defined networks, telecoms edge compute and AI-enabled network automation, have their basis in Linux Foundation projects.
The increasing participation of companies from IT backgrounds in the heart of telecoms networks helps to create confidence in open source technologies, which have been a key element of IT and data center platforms for decades. Intel recently took over the development team of the Open Networking Foundation (ONF – see below), increasing its role in a very important contributor to open telecoms systems, including open source implementations of the mobile core and the RAN Intelligent Controller (RIC).
As large IT players such as HPE and Juniper pile into strategic pieces of software like the RIC, we are likely to see more start-ups turning to open source communities to accelerate their own activities and stay included in the ecosystem.
And where the hyperscalers go, especially Google, there is usually a strong appreciation of how open source communities can foster innovation and accelerate change – and Google, like IBM before it during the IT industry’s open source revolution, is a master of steering open efforts and then turning them to its own advantage, as it has done with Android.
Google’s latest cooperation with an open source project is Nephio (see below), a project with Linux Foundation to create cloud-native automation and management of 5G networks across multiple edge locations. This description could hardly tick more boxes in terms of operators’ current buzzwords and strategic architectural concerns.
But open source models are immature in telecoms networks and have their trade-offs, not least the threat of fragmentation if a giant vendor or two does not take a lead. Not all valuable open initiatives need to be open source – the emergence of real specs from OpenRF, which we cover today, is an example of this, and has the potential to boost supplier diversity in this critical technology, without going the whole way into open source.
This is likely to be the case in most of the RAN, with the possible exception of the RIC, if the ONF’s SD-RAN becomes powerful.
The ‘open source RAN’ remains the stuff of dreams, even according to some of the challenger vendors that are making progress in opening up the cellular ecosystem. Some of this relates to the difficulty of applying classical open source models, as epitomized by the frameworks of the Apache Foundation, to chips and other hardware. But even when it comes to the software element of the virtualized RAN (vRAN), there is very little work being done on RAN network functions that could be downloaded for free to any baseband or radio unit.
For those that believe a true open source model is the best way to throw open the doors to the RAN market, a depressing sight is that of the WiFi sector. The WiFi ecosystem is undoubtedly more open than that of cellular, with a diverse set of vendors of access points and other elements. The openness has helped make WiFi equipment affordable and applicable to a huge range of use cases. But the IEEE and WiFi Alliance, which publish the specifications, are conventional standards organizations; open source efforts such as OpenWrt have been fragmented in their impact; and in many higher end systems, proprietary embedded software remains the norm.
Of course there is plenty of open source code in use in mobile networks, not least in various flavors of the Linux operating system and in elements of NFV and containerisation infrastructure. But so far, opening up the RAN to a broad base of software developers is being achieved through reference platforms rather than pure open source platforms.
These may be defined and distributed by open organizations such as the ONF, or by individual vendors that open up their platforms in order to achieve broad impact for their products and services. An example of the latter is Intel’s FlexRAN software reference stack, which has been extensively used in early vRAN developments and trials (and as noted above, Intel itself is now getting closer to the ONF).
Some argue that the only way that radio software – of sufficient capability to win operator trust for their macro networks – would be made available under a general open source licence would be if a giant like Intel triggered the move. If Intel open sourced FlexRAN, it would be gambling on achieving the kind of dominance of the ecosystem that IBM did with some of the IT technologies it committed to open source in the 1990s.
But it might just split the industry – after all, AT&T put its xRAN code into open source, but while that was a catalyst for O-RAN, that has moved away from general open source licensing. And IBM was, though wounded, still the king of IT back in those days, so its action was more akin to Ericsson open sourcing its radio platform than Intel. A hypothetical open FlexRAN would still need the big RAN vendors on board, just as O-RAN will to achieve maximum scale, and we are still waiting to see the outcome of the game of ‘who blinks first’ between the operators and the largest vendors (with Nokia playing both sides).
In an interview last year with SDxCentral, Steve Papa, CEO of O-RAN vendor Parallel Wireless, said: “Maybe open source in radio is all about the chicken and the egg, and someone like Intel needs to step up and catalyze it. Otherwise all you’re doing is creating proprietary stovepipes. FlexRAN, if it’s only for Intel processors, well someone else is going to build one for other processors. And the operators lose; in fact, open RAN moves slower, moves slower for Intel as well.”
He added: “We need people that want to accelerate open RAN to step up and accelerate it by contributing some of this stuff to open source for the radio.”
For now, ‘open’ organizations focused on the RAN are adopting a hybrid approach. ONF, for instance, uses a combination of elements, some defined by open source organizations and some by standards bodies. Some of its platforms, such as the Open Network Operating System (ONOS) are fully open source and subject to the Apache process. Others rely on licensed technologies. It offers its developments under two main models:
- Technical specifications include ONF’s reference designs and OpenFlow portfolio of standards, and are licensed under the Foundation’s RAND-Z (reasonable and non-discriminatory) IPR policy for licensing.
- Technical recommendations, which are normative documents that define APIs, data models, protocols, information models and similar. These are open source under the Apache 2.0 system and may be freely used if formally cited, not altered, and not commercially sold.
When Telecom Infra Project (TIP) was first established by Facebook, it was billed as an open source initiative to drive commonality and scale economics into the telco network. But it became clear that a more nuanced approach was needed to attract companies that had developed highly differentiated contributions. Many small vendors – as well as many operators – remain concerned by the risks that open source brings in terms of quality, control and patent licensing.
One of the start-ups that was chosen, in 2017, to be incubated by Orange under the TIP ecosystem acceleration center (EAC) program was Amarisoft, whose CEO Franck Spinelli said at the time: “We don’t want to let our technology go for nothing. We want money for that.” While Facebook (and others) have open sourced their own contributions to the project, the option of licensing under RAND (reasonable and non-discriminatory) principles is open to any working group or member, and in 2017, Facebook initiated a new TIP group specifically to support RAN licensing.