IS-Wireless: Massive MIMO is 'a battlefield for closed RAN'

Microwave ovens are obviously intended for heating up yesterday's leftover curry, but what if the processors inside them could also be used to run part of a radio access network (RAN)? It's an extreme example of reusing existing compute resources to create super-densified networks of the future, and the vision of Sławomir Pietrzyk, the CEO of a Polish startup called IS-Wireless.

Pietrzyk thinks the mobile telecom industry should be prioritizing such network densification over concepts like massive MIMO, an antenna-rich 5G technology championed by the big kit suppliers. "Traditional vendors are very interested in cementing their status quo by promoting solutions that are hardware-oriented," he told Light Reading.

"I am very skeptical about every proposal a traditional vendor puts in," said Pietrzyk. "To me, it is a trap, and the open RAN wave-one guys got trapped on a battlefield prepared for somebody else. This is a battlefield for closed RAN."

He is, of course, referring to a recent Ericsson-led proposal for an uplink modification to open fronthaul 7.2x, the specification designed to support vendor interoperability between radio units (RUs) and distributed units (DUs) in open RAN infrastructure. A compromise between that proposal (backed by AMD, Ericsson, Nokia and ZTE) and another one from AT&T, Orange and Qualcomm has now been approved by the O-RAN Alliance, the group developing open RAN specs.

But it is not universally popular. Ericsson argues that open fronthaul 7.2x is ill-suited to massive MIMO because too many critical components reside in the DU. But moving some of these back into the RU, as it proposes, threatens to add complexity and hinder open RAN progress.

South Korea's Samsung abstained from voting out of concern about fragmenting the ecosystem. Parallel Wireless, a small open RAN software developer based in the US, did likewise. And in an email sent to Light Reading, Japan's Rakuten acknowledged there had been "discussions around the downsides of making the O-RU [open radio unit] more complex with additional electronic processing demands on O&M, field maintenance costs, etc."

Growing recognition

IS-Wireless is by far the smallest player so far to have voiced objections to the proposal for a new spec. Headquartered near the Polish capital of Warsaw, it employs about 50 people and has raised about US$8 million in funds until now, according to the Crunchbase website. But it has carved out relationships with some big ecosystem players and demands to be heard.

That's largely because when Europe's big five telcos – Deutsche Telekom, Orange, Telecom Italia, Telefónica and Vodafone – published a white paper in November 2021, calling for a stronger European open RAN ecosystem, IS-Wireless was mentioned as the only small European software player with a complete portfolio of products.

Any rational observer might assume a startup working in an area the telcos deem so critical would have attracted funding from the telcos' own venture-capital arms. But that has not been the case. "Regarding funding, the capital supply side actually got very difficult," after late 2021, said Pietrzyk. "We were looking for a round of $10 million but in the meantime the war broke out [in Ukraine] and that changed a lot."

Instead, IS-Wireless has had to fund development through revenues. The good news is that sales for 2023 are already well above what the company made in 2022 and doubling from quarter to quarter, according to Pietrzyk, although he would not cite numbers. IS-Wireless is still seeking investment, but the need for it seems less critical. Any funds it does manage to raise would be used to bridge some efforts, said Pietrzyk.

If the big telcos' report did not result in new funding, it has generated some attention for IS-Wireless and led to a few significant deals. Among those is a contract with Werner von Siemens in Germany for the deployment of a small network in Berlin. Other companies involved in that project are Osram, a maker of electric lights; Fraunhofer, a German research institute; and T-Systems, the IT part of Deutsche Telekom and a company Pietrzyk evidently regards as an important partner.

Appetite for densification

What marks IS-Wireless out from all the other open RAN developers out there? Apart from its Europeanness, Pietrzyk makes some bold claims for his firm's software. The competitors he derides as "wave-one" players have often come from a 4G background, and they have only gone so far when it comes to disaggregation, he argues.

"The protocol stack for the RAN, even if split into DU and CU, is still blocky. It is still a monolith that you port on a server that needs to be dedicated and that needs to sit somewhere on the site," he said. "That architecture has not been changed over decades. It is part of the traditional RAN, and it is kind of followed by the open RAN wave-one vendors."

IS-Wireless claims to have a software stack – covering all the various Layer 1 to Layer 3 functions – that can be disaggregated into much smaller elements than just the DU and the CU. This essentially means a DU could be split across several hardware devices, including compute infrastructure currently in the field. While Pietrzyk's example of the microwave oven processor might be a colorful exaggeration, IS-Wireless was able to put its RAN on a customer server in Krakow already supporting other applications, said Pietrzyk.

He is unprepared to share many technical details of the software at this stage, noting that IS-Wireless has filed a patent for the technology. But he uses words such as "fluid" and "liquid" to describe the general concept. "We granularize and split the software stack into many small elements," he said. "It is a sort of fluidity or liquidity in a software stack – a liquid you pour into a form and when it gets frozen you have your solution available."

Betting on the future

All this begs many questions still unanswered at this stage. Pietrzyk concedes openness has limits and that separation of software from hardware is much harder at Layer 1, the part of the stack responsible for computationally intensive baseband processing. This seems partly to explain his aversion to massive MIMO. "If you want to be portable and have widely available systems, you should simplify things that were complex so far."

His vision of a world where edge networks are pervasive and compute infrastructure is ubiquitous seems a long way off, too. Massive MIMO appears to be going mainstream and shows up frequently in telco RFPs (requests for proposal) to vendors. Most operators in Europe have, at the same time, done little or nothing on network densification.

But Pietrzyk is clearly excited by the rise of neutral hosts such as Cellnex, an infrastructure company that rents space on towers and network equipment to retail operators. And he expects to see even more of that in the next few years. "I would not be surprised if owners of non-telco infrastructure like cities and petrol stations and kiosks get involved because they have a very important asset, and that is localization," he said. "The physics will not change, and we need to densify."

To support its go-to-market strategy, IS-Wireless has formed partnerships with a wide variety of companies, from Indian systems integrators to Taiwanese hardware producers. But it really needs companies investing in RAN infrastructure to share Pietrzyk's vision and decide his software can help them realize it.

His hope is that the market will look very different to operators by the mid-2020s, when contracts signed in recent years have run their course. "Before that, their hands are bound because they are in a vendor lock-in situation with Ericsson, Nokia and Huawei," he said. The next few years will show whether he's right.

Related posts:

— Iain Morris, International Editor, Light Reading