Your browser is not supported. Please update it.

16 May 2016

Mobile giants will not be able to hold back the open source wave much longer

The traditional core of the mobile industry still runs shy of open source. Qualcomm may have made some concessions to the new world with activities like AllSeen, but in general, mobile technologies are still run by traditional standards bodies and industry alliances with complex patent sharing deals and strictly controlled development processes.

But the new giants of the mobile world have a very different view of open source, seeing it as a way to drive innovation, accelerate development and share costs. Sophisticated players like Google are expert in making their inventions open, to gain the benefits of a broad ecosystem and eager developers’ efforts, while retaining a tight level of control over the roadmap (though that control has slipped somewhat in Android).

Increasingly, as networks morph into IT platforms and operators adopt virtualization, they will be dealing with new vendors – ones which thrive on the open source model. There are many downsides to it, of course – the battle over management and orchestration standards for NFV is a good example, with the open source options fragmented and with the lead candidate, OpenStack, coming with compromises in terms of carrier-grade functionality.

But as previously seen in the enterprise, open source will take over in the end. Its cost and innovation benefits are just too attractive to resist, and role models like IBM will show the mobile world’s giants how to support the trend while retaining control over their ecosystems and platforms.

And if the network vendors think this is an IT and apps layer issue, and not one to touch the rarefied technologies of the network itself, they are wrong. There were three open source announcements this week, from the three giants of the web – Facebook, Alphabet and Amazon. Amazon’s related to deep learning, a key technology for next generation web experiences. Alphabet’s was the open source release of the Thread protocol, courtesy of its Nest Labs subsidiary. But more worryingly for the traditional networks community, Facebook was placing a whole software-based routing technology – the kind of concept which could derail the conventional network suppliers – into the open community.

Facebook, like Google, has been working on a wide range of technologies, including drones and satellites, which can extend broadband access to the entire world, and improve the cost and quality of that connectivity, to drive further usage of the company’s web services. The social network giant has unveiled a new software routing system for the physical network it announced at its F8 developer conference last month.

This consists of the Terragraph gigabit radio, for building low cost cellular and WiFi small cells in dense urban environments; and a concept design for a Massive MIMO macrocell for rural areas, called Project Aries. These are now complemented by the Open/R software routing system, which has been developed to direct traffic from Terragraph and Aries – though Facebook says it has also been adapted to work with more commercially real parts of its networking infrastructure.

Terragraph runs in 60 GHz spectrum and, as well as Open/R, features “SDN-like” cloud controllers to handle large numbers of cells efficiently and flexibly, targeting capacity where it is required at any one time. The system can support multi-Gbps links and is IPv6-only. It will be trialled in San Jose, California later this year.

Meanwhile, the Project Aries (Antenna Radio Integration for Efficiency in Spectrum), prototype supports 96 antennas and 24 data streams, targeting spectral efficiency of 100bits/sec/Hz. So far, Facebook has demonstrated Aries at 71bps/Hz.

Not only is Facebook impinging on the world of Nokia and Ericsson, but it is not keeping its advances shrouded in patent-protected secrecy as they do. The web giant says it has reused as much open source code as possible to accelerate development of its technology and tap into others’ innovations, and is now working out the best way to introduce Open/R to the open source community for broader use and enhancement. It will also be part of the Telecom Infra Project (TIP), which Facebook set up in February, working with a group of operators to open source infrastructure hardware and drive down costs with open commodity platforms. Terragraph is also being contributed to that process.

“Open/R makes it possible to prototype and deploy new network applications much more quickly than with the industry’s standard development process,” Facebook said.

Routing protocols like Open/R will be essential as wireless networks move into high frequency bands such as the licence-exempt 60 GHz, where Terragraph lives. A smart protocol helps to find the best signal path in spectrum which is poor at penetrating walls or foliage, and is sensitive to atmospheric conditions.

“Open/R was originally designed as a shortest-path routing system to power Terragraph, our multinode wireless network that delivers high speed internet connectivity in dense urban areas via short hop transmission between small nodes,” Facebook wrote in a blog post.
The combination of millimeter wave spectrum, smart routing, gigabit radios and Massive MIMO look as close to a proto-5G system as most of the demonstrations coming out of major operator and network vendor labs – virt Facebook is also working on the user experience to go with these advanced networks, full of virtual reality and AI-driven interactions. Most worryingly for traditional vendors, it certainly takes the view that 5G will not just be about architectures but about new ways of procuring and deploying equipment – standardized, commoditized, open source platforms will dramatically change the conventional model.

By open sourcing many of its designs; developing alternatives to fiber for remote area backhaul; pushing for free or flexible spectrum such as white spaces; and working with non-traditional operators or infrastructure partners, Facebook has the opportunity to rewrite the rules for how wireless networks are built and costed, driving down the capex and opex bills and lowering barriers to entry for innovative service providers which can make a business case for targeting the unserved billions (something which is hard to do with current architectures, spectrum and revenue expectations).

Jay Parikh, VP of engineering at Facebook, said the firm’s Connectivity Lab unit is interested in “radical new approaches to get the connected unconnected … Our rule in the Connectivity Lab is we’re looking for gains that will make things 10x faster or 10x cheaper or both.”

Facebook’s activities, like Google’s, are a double-edged sword for MNOs. On the one hand, these web giants threaten to lower the barriers to entry for new and agile service providers; on the other, they are doing much of the heavy lifting to create a new generation of equipment which will transform the economics for existing MNOs, allowing them to build on their market advantages of brand and network experience while expanding those networks far more cheaply, as seen in the Telecom Infra Project.

Many of these developments are mirrored at Google parent Alphabet, and the latest in the firm’s string of open source contribution is OpenThread, a free open source version of the Thread protocol which was developed by smart home subsidiary Nest, and is being promoted as a de facto standard by the Thread Alliance, formed last year. The protocol is a strategic one to Alphabet, since it has the potential to squeeze Bluetooth, ZigBee and Z-Wave in the smart home personal area network space – allowing the web giant to control the direction of standards in this key market, and to develop services that can harness smart home data.

Thread supports basis in IPv6 and is an 802.15.4 protocol akin to 6LoWPAN. It has been edging closer to ZigBee (ZigBee, along with Google Weave, can run on top of Thread), and Nest has been clear it would like the two similar platforms to converge, uniting against Bluetooth LE, which is playing catch-up in some areas, notably mesh networking.  Thread, whose Alliance co-founders were ARM and Samsung, is making progress with chip availability – Atmel, Dialog, Qualcomm and Texas Instruments are contributing to its development and there are development kits from NXP and Silicon Labs.

A full open source version could accelerate innovation and uptake. OpenThread will be distributed on GitHub for free use based on a BSD licence. It can run on home routers with 64KByte RAM and 256KByte ROM and end devices with less memory. Thread already runs on Nest’s products and Google’s OnHub home router and the Alliance says over 30 other products are waiting for certification. The Alliance now has over 230 members.

Over at Amazon, the retailer’s latest open source development is in the area of deep learning, which will power many web and mobile services of the future. It has open sourced its Deep Scalable Sparse Tensor Network Engine (DSSTNE, pronounced Destiny) library on GitHub under the Apache 2.0 licence.

Currently, Destiny is harnessed to enhance Amazon’s core business, helping online shoppers to find the right product from the firm’s vast catalog using neural networks. But the engine can clearly be used for a far wider range of applications, either by Amazon itself – as it expands its presence across the whole digital experience – or by third parties. It said it was releasing the software in open source “so that the promise of deep learning can extend beyond speech and language understanding and object recognition to other areas such as search and recommendations. We hope that researchers around the world can collaborate to improve it. But more importantly, we hope that it spurs innovation in many more areas.”

Amazon says DSSTNE is different from other deep learning libraries because it is optimized to work where there is only sparse data available, and because it is very fast – more than twice as fast as Google’s TensorFlow open source machine learning platform, claims the retail giant. Something of a battle to entice developers to their AI engines is developing – Google open sourced TensorFlow back in November and Facebook also offered its machine learning and AI tools to the community early last year. And in December, Amazon Web Services joined the open source, non-profit AI research project, OpenAI.