Forecasters continue to bicker about the sheer size of the ‘internet of things’. Ericsson’s 50bn connected devices by 2020 was probably just meant to wake attendees up at a conference, but has become almost received wisdom. Researchers at IDC are thinking even bigger though, predicting 212bn connected ‘things’ in the same timeframe, generating revenues totalling $8.9 trillion for services, devices and surrounding infrastructure.
IDC is taking a very broad view of IoT ‘ it calculates that this already generated spend of $4.8 trillion last year, when the term was scarcely heard of. Of course, there are indeed many machine-to-machine systems already in place ‘ the big changes over the coming years will be the number of endpoints connected, as the market moves beyond telematics to penetrate all areas of life; the need for broadband connections for some M2M apps; and the increasing intelligence of some classes of ‘things’ as they exchange information with one another and the cloud. By 2020, IDC thinks 30.1bn of its 212bn connected objects and sensors will be autonomous and intelligent, supporting trends such as smart cities, self-driving cars and intelligent energy.
The timeframe does seem ambitious, given the immature state of standards and the ecosystem. Much of it is currently based around traditional applications with specialized technology, such as fleet tracking or industrial control; or around a few services that have achieved backing from governments or large companies, such as smart metering. Of course, an immature ecosystem provides the best opportunities for companies to push themselves forward and grab the rudder.
Chip companies are racing to the space, especially those ‘ all the way from Intel to specialists ‘ which are squeezed in the mobile market by the dominance of Qualcomm. Operators are desperate to carve out a profitable role in the value chain, rather than just becoming bitpipes for the huge amounts of data generated by IoT, and a few are making real progress in devising whole service offerings, as well as stimulating the device ecosystem. And the IoT, given its big data implications, is an even better opportunity than the mobile data boom for enterprise platform providers and integrators to leap the chasm into the telecoms space, so we will increasingly see IT players battling with carrier vendors ‘ Cisco versus Alcatel-Lucent for the routers, for instance, or Oracle versus Ericsson for the back end systems.
On the chip side, many of the companies which fled from the hugely competitive and consolidating smartphone market are now adapting many of their low power, wireless connected, system-on-chip developments for IoT sectors such as smart cars or even ultra-low power sensors. Of course, the list of smartphone wannabes also includes Intel, which this week revealed its full roadmap for IoT, as it bids to fend off ARM and avoid repeating mobile history in the embedded field.
There are four legs to the firm’s IoT strategy ‘ to extend the benefits to legacy systems; to help back end IT systems scale up; to enable data analytics and security at device and cloud levels; and to provide a common framework across networks and devices, courtesy of x86 and its Intelligent Device Platform, based on software from its Wind River real time Linux unit.
As previewed at the giant’s developer conference last month, a combination of low power Atoms and a new family called Quark will address connected and embedded devices from next year. The first offerings are the Atom E3800 processor, codenamed Bay Trail-I, and the ultra-low power Quark X1000 SoC. Bay Trail-I’s key initial targets are portable medical devices, in-car infotainment, point of sale terminals and kiosks, as well as gateways for connecting IoT sensors and low end or legacy endpoints.
Initial Quark targets are those endpoints themselves, particularly in the industrial, energy and transportation segments, as well as the emerging wearables segment (CEO Brian Krzanich showed off a prototype connected bracelet when he announced the platform at the developer forum).
Last week, Intel announced that it was also incorporating Quark in a small computer called Galileo, based on the Arduino open source hardware and targeted at do-it-yourself users, who are expected to contribute to the proliferation of IoT objects by creating small, simple, open devices for all kinds of functions.
Bay Trail’s role is clear and fairly high end in IoT terms. It basically follows through on previous Intel promises to squeeze Atom into smaller, less power hungry formats, suitable for gateways and larger connected devices.
But with Quark, Intel is trying to carve out a segment which it can define and control, and that makes it a bigger gamble. This segment sits somewhere between a connected device that can run a full CPU and operating system, and one that runs a basic microcontroller. The latter category dominates the existing M2M space, and ARM is powerful here with its M0+ Flycatcher core. Microcontroller-based SoCs are extremely small and power efficient and fall into the ‘silicon dust’ price range of less than 50 cents.
Intel is banking on a middle market opening up between these two established categories, for devices which run an OS (embedded Linux) and CPU. It believes the rise of intelligence M2M will drive the need for these more functional SoCs, which will therefore be able to bear slightly higher prices, power levels and size (and could also, in future multicore versions, be applied to low cost handsets). To address that hoped-for need, Quark is a whole new design ‘ not down at microcontroller level, but promising just 20% of the size of Atom and 10% of the power consumption, and with an ‘open architecture’.
This last feature is the most interesting, but also the least precisely defined by the chip giant. It seems it will be prepared to open up elements of the platform to others, recognizing that for the hugely diverse internet of things, one vendor will not make every type of chip required. It will do better to sit like a controlling spider at the center of a web of fabless semiconductor firms and associated software developers, as ARM does in the handset world. Not that Intel is licensing the core itself, at least not yet. Instead, it will offer ‘fully synthesizable’ chips with hooks in the silicon to add others’ IP blocks, to customize it for specific tasks. It also unveiled a software stack that includes security, manageability and connectivity features optimized for IoT.
At this week’s official launch, Intel also announced intelligent gateways which will handle the process of connecting existing embedded products that are not linked to the internet while adding security and analytics functions. Intel says that 85% of the embedded base is “built into legacy” and needs to be connected.
Intel’s two major software acquisitions of recent times, of security firm McAfee and real time Linux provider Wind River, are coming into their own around Quark as they provide the data processing, intelligence and protection that a modern connected ‘thing’ requires. Their software will be embedded at chip level. Wind River is critical to Intel’s promise to deliver an integrated services framework from device to cloud. Its role is to ‘deliver the intelligence in the internet of things’ with messaging and development environments geared to embedded devices, as delivered by its upgraded Intelligent Device Platform. This uses the MQTT (Message Queuing Telemetry Transport) protocol for messaging between IoT and also supports communication over Wi-Fi, Bluetooth, ZigBee and other short range wireless protocols which are fighting for their place in this huge market. It also supports applications development using Lua, Java, and OSGi.
Also key to Intel’s ability to differentiate itself in this field is the integration of McAfee’s security on-chip, which the firm has been talking up for some time. Last year, McAfee’s CTO Stuart McClure spoke of the need to extend security features to embedded devices, from phones to M2M modules, increasingly at chip level. He argued that it was untrue that embedded gadgets could not be secured, and this should be addressed proactively by the industry, rather than waiting for disasters to happen.
“It is an invisible world that we don’t see all that much, but these are not something you can push from the cloud and update,” he said. “We need to think about this world equally along the desktops and devices.” That has certainly materialized in the Atom/Quark platform which are seeing Intel ‘push the protective and diagnostic technology all the way down the stack. The threat is moving further and further down, we have to push these features further and further into the firmware, the bios and the silicon itself.”
Ton Steenman, VP and general manager of Intel’s Intelligence Systems group, highlighted the opportunities for Intel at both ends of the IoT spectrum ‘ the devices, but also the servers to support the vast amount of analytics that will make sense of all this data.
The IoT is becoming real because “big data has matured in a significant way, matured by business intelligence and social media. Without those critical insights, you will be unable to drive the real business potential that the industry is after,” he told the launch event, quoting McKinsey research that said connected devices had grown at a rate of 300% over the past five years.
The creation of a hardware and software platform which spans the back end and the device has already scored a significant win for Intel which, along with Cisco, this week became a partner in the important ‘Industrial Internet’ data analytics platform created by General Electric. GE also anointed AT&T as king carrier in this fledgling market by including it as the only carrier partner.
During the two-year life of its Industrial Internet program, GE has expanded it from providing services, such as energy management and remote diagnostics, to its own customers in key sectors like energy. It has now turned it into a huge IoT framework in the cloud, into which, in future, a wide range of service providers, developers and devices can integrate, well beyond GE’s own client base. This will revolve around Predix, its platform for industrial and M2M apps which can be created for vertical market ‘things’ from aeroplane engines to smart meters, and remotely managed. Predix has four elements to its framework, all based on standards ‘ sensors, analytics, device management and developer APIs.
GE is adding 14 more data predictive capabilities to its cloud-agnostic service, bringing the total to 24 since it unveiled the project in June, claiming to be providing the first data analytics platform that can manage massive scale industrial machines in the cloud. Many of the new functions are geared to the aviation, healthcare and energy sectors, all frontrunners in IoT use. For instance, it has added Cloud Imaging for healthcare.
GE’s partners are not just supporters on a press release ‘ each has committed to a significant contribution to the platform, although the ecosystem will open up to third party developers and a wider section of partners next year.
First wave participants included Amazon Web Services on the cloud side, Accenture and EMC’s Pivotal unit for data analytics and tools. Among the new sign-ups, Cisco is working on collaboration and analytics services between machines, based on open standards.
Intel will work with GE to embed virtualization and standardized, cloud-based interfaces within Predix. This work, in turn, will help move towards an open infrastructure to which global service providers can connect. Bill Ruh, VP of GE’s software and analytics center, told GigaOM that he liked Intel’s IoT message about embedding intelligence from the server to the edge using common tools. ‘We’ve used Intel in our machines for a long time,’ he said. ‘Quark is a great product and we saw that, and realized that is an opportunity to leapfrog to the next generation. The world is changing and everything will need to have intelligence in it, even at substandard-size.’
AT&T has perhaps scored the biggest win, since it will provide the whole wireless and wireline communications system for Industrial Internet, and will enable GE machines to connect to AT&T’s cloud for remote monitoring, diagnostics, and maintenance. This is a significant boost to AT&T’s reputation as one of the top three operators in the fledgling connected M2M market, both in terms of the services it supports across its networks, and the work on endpoints which it drives through its Connected Devices program. There is also a major revenue opportunity – it is unclear at this point whether GE customers will buy AT&T’s connectivity separately or bundled in, and how it will be priced, but the number of devices involved could be massive once the platform gathers momentum.
Morgan Mullooly, an analyst at Analysys Mason, commented: ‘This is a hugely significant win for AT&T. We expect a tremendous number of M2M connections to be activated in the next 2’3 years, as millions of industrial components roll off GE production lines fitted with embedded M2M modules and are dispersed around the globe. Consequently, we expect the deal will strengthen AT&T’s position as the leading US M2M service provider by number of connections.’
Other major industrial device and equipment makers will follow GE’s lead, adding value and revenue streams to their basic business. Operators will have a similar opportunity, by harnessing their embedded SIM cards to make themselves the central point for integrating the many elements of the ecosystem. As in smartphones, this will become a battle with the device suppliers to control the framework and the overall experience, but embedded device firms will have very different agendas to those focused on the consumer or enterprise. If carriers take the GE/AT&T collaboration as a template, supporting open standards and developing common platforms, they could see the IoTopening up as a growth sector where they work hand-in-hand with device suppliers, rather than against them, and share the pot of gold which beckons.