Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.
July 18, 2013
16 Min Read
InformationWeek Green - July 22, 2013
Download the entire July 22, 2013, issue of InformationWeek, distributed in an all-digital format (registration required).
Here comes the internet of things
If you liked the early days of cloud computing, you're going to love the Internet of things (IoT) and its less-sexy cousin, machine-to-machine communications. Certainly, you'll be in elite company. Cisco is dedicating an entirely new business unit to the fledgling effort. AT&T has built two shiny new facilities dedicated to developing things like smart luggage that can locate your bags in the airport so you don't lose them. Verizon has a program aimed at transportation. Broadcom, Oracle, Samsung -- all are in the hunt. Intel says IoT technology will enable 3.8 billion more connected "things" by 2015. At an average cost of $100 per item, we're talking $380 billion (about the GDP of Austria) in just two years.
IoT fever isn't limited to IT vendors. General Electric, the 121-year-old multinational conglomerate not exactly associated with the build-your-own-market-bubble crowd, says its IoT offerings, under the rubric "Industrial Internet," will deliver at least a 1% improvement in efficiency to customers by way of sensor-enabled industrial equipment. Total value of that 1%? To start, $66 billion for its energy customers, $30 billion for aviation and $63 billion for healthcare.
Sounds like something enterprise IT should get in on, right? The problem is that IoT is still a mix of ill-defined terms, aspirational concepts and extreme hype (see "early days of cloud computing"). There are solid examples of the benefits of pervasive interconnectedness, but they simply don't yet merit the stratospheric projections we're seeing. To wit, Cisco pegs what it calls the "Internet of Everything" as delivering $14.4 trillion of economic value by 2022. To put that number into perspective, the total U.S. gross domestic product in 2011 was $14.9 trillion. The hole blown in the world economy by the financial industry meltdown of 2009 is generally agreed to have been about $16 trillion, and the entire debt of the U.S. government at publication time was $16.7 trillion, give or take.
Way to think big, Cisco!
To be fair, the company is putting its R&D money where its mouth is, although a dedicated unit of 500 people, mostly shifted from other company divisions, with a $200 million budget seems on the modest side for technology expected to increase asset utilization by $2.5 trillion, up employee productivity by another $2.5 trillion, boost supply chain and logistics efficiency by $2.7 trillion, improve the customers' experience (and thus their loyalty and spending) by $3.7 trillion. The remaining $3.0 trillion Cisco allocates to "innovation."
Now that we have your attention, contemplate the changes that must happen on everything from assembly lines to IPv6-ifying networks to data management, security and even public policy to get those kinds of returns. It's just damn heroic, like a Marvel comic hero but with APIs instead of an arc reactor.
The really crazy thing is that while there's some of what Alan Greenspan called "irrational exuberance" in these IoT numbers, in the long term -- probably longer than Cisco's 2022 timeline -- projections like these may be only a bit overwrought. That's mainly because the IoT takes balanced advantage of Moore's Law.
For the most part, Moore's Law so far has delivered a significant increase in computing performance paired with a lopsidedly modest decrease in system prices. The $5,000 desktop PC of the '80s now costs $500, a 90% drop, but the CPUs powering today's computers are about 10,000 times as powerful as when Intel 8088 processors roamed the Earth.
The IoT will emphasize technology that more evenly leverages Moore's Law -- performance increases will come with commensurately substantial size decreases and resulting power and cost reductions. A reasonably powerful 32-bit computer with a thin-film battery and wireless network radio can be built for a few dollars. As a result, for example, smart-home technology is starting to become reasonably priced. A smart outlet now costs about $50, compared with $5 or less for a dumb outlet without any instrumentation. The 10-1 price ratio is still enough to give most people pause; going smart will add about $2,500 to the price of an average house just for the outlets, and it will take a long time to earn that back in power savings. But once the ratio is more like 3-to-2, smart outlets will become the norm.
Report CoverOur report on making the transition to IPv6 is free with registration. This report includes 12 pages of action-oriented analysis.
What you'll find:
How to bring IPv6 to the desktop and support remote offices and telecommuters
Steps for upgrading systems, websites and enterprise apps.
Unfortunately, that's only a sliver of the IoT picture, because the technological hurdles will be the easy ones to clear. Hardware (versus software, which we'll get to) is almost a solved problem. In our outlet case, manufacturing the computer components isn't the costly part. The expense comes when you build those electronics into the outlet and add the high-power control circuitry to actually turn the outlet on and off in response to whatever triggers the homeowner sets.
This is something of a universal issue for the IoT. In the world of network-connected sensors -- which are key to not just smart outlets, but everything from health monitoring to traffic management to more-efficient manufacturing processes -- it typically costs about five times more to put the sensor in place than it does to manufacture it. That implementation cost must come down before most existing machinery will be retrofitted. Unlike cars and computers, which have four- to 10-year lifespans, we can't wait for industrial machines to hit replacement age, which can often be 20 years or more. Thorny issues like ethics and privacy must be worked through as well.
That's not to say you should ignore IoT technology. Some technologies will be slam dunks and pay off sooner rather than later. Sectors such as pharma, transportation/logistics and government already are benefiting from IoT advances (examples below), and if you're in one of these sectors, you'd better be paying close attention. If you're in a different industry, start thinking about how your company can leverage this meeting of big data, mobility, business process and the cloud.
Here are some more examples to get you started.
IBM's Counterfeit Drug Battle
For some time, IBM has worked on keeping counterfeit medications from sneaking into the supply chain. Its goal is to ID drugs right down to the individual pill using a variation of the microchip IDs that have been around for a while (you can get one inserted into your pet for about $10). The plan is to bring that cost down 100-fold, so that individual pills or single-use injection vials could be tagged for about 10 cents. Drugmakers simply drop a chip about the size of a grain of sand and coated with a nonreactive substance into the vial or into the pill as it's being pressed. The objective is to go from a rate of about 10% fakes today to far less than 1% in the future. Bad batches can also be identified and removed from the system easily.
Logistics and transportation are obvious places where Internet-connected sensors can help. If UPS can accurately predict an imminent truck breakdown, other trucks can be assigned to meet the vehicle with the problem and avoid missing a delivery guarantee. Route optimization can also be improved. UPS famously instituted a no-left-turn policy for its vehicles to speed up deliveries and save fuel. With sufficient real-time traffic information, companies can make route planning even more efficient, and that's just the beginning of what the IoT can do. Chips can limit the top speeds of service vehicles, for example, and report deviations from routes.
Then there's the promise of smart homes, factories and offices, which through design, instrumentation and automation can substantially lower energy usage and costs. Still, while it may run just $30 to $50 to network-connect an outlet, motorizing and networking blinds and air ducts is much more expensive, often $200 to $500 a pop. It's not uncommon for that work to tack on $30,000 to $50,000 to the price of a typical house.
Clearly, many IoT technologies have a distinctly first-world bent. That's fine for now, but it's not likely that developed countries are where billions of new connected devices will come from, never mind trillions in savings. For that, we need universal connectedness and problem-solving on a global scale. Let's look at some of the factors holding us back.
Evolving The Internet
In almost all IoT applications, sensors on local networks report back to a single entity that then connects to the Web and delivers data upstream. In the home, for example, wireless standards such as Z-Wave exist for operation in the 900-MHz band with a range of about 100 feet and a maximum of 232 devices. These local islands will connect to the Web through a central controller. So while your refrigerator may know you're out of milk, chances are it'll be some other, smarter device that actually calls the grocer for a refill. Thus, the smart home won't cause as much of an explosion of devices on the Web as if each refrigerator, toaster, dryer and DVR were online.
The same is true for industrial applications, where it may not be every valve along a pipeline (or whatever application you can imagine) that's hitting the Web. Still, even at a 10-1 or 100-1 ratio, there will be a lot of new devices to contend with. Even the addition of a half billion or so home-based controllers could have significant impact as they interact with the power grid, surveillance and security firms, and other businesses. IP addressing will be an issue -- IPv6 finally has a killer application -- but the added traffic must also be managed.
Still, what's really going to be important are the security, privacy and process algorithms that will deal with data from potentially millions of sources, some of them reliable, some of them not.
iPv6 by the numbers
In terms of connecting humans to the IoT, consider the Nike FuelBand. The idea is that you wear the device and it tracks your activity throughout the day, to monitor your caloric output. Unfortunately, that's at least a little bit of hype -- version one is nothing more than a pedometer, a clock and a Bluetooth radio. But the technology will surely improve, and the device comes with some pretty slick software that lets you compete against yourself or others and post your fuel points burned to Facebook. Nike itself uses fuel points to reward its employees -- it might announce at 9 a.m. that the first to show up at reception with 3,000 fuel points wins a pair of Portland Trail Blazer tickets, for example. Apple's upcoming iWatch will likely have similar functionality.
Beyond commercial products, specialized wearable monitors will be a boon for medicine. They may be able to spot changing conditions in high-risk patients and predict critical events before they happen. Short-range communications such as Bluetooth will transmit data to a local device that can do some autonomous diagnosis and report back to a hospital via a cellular data connection. Could such a smart patch simply talk to your Android or iPhone, which could notify your doctor if it senses a problem? Sure. The FDA may have something to say about it, but the idea is the same.
Really Big Data And Iffy Connectivity
An even bigger challenge than hardware will be the software on the edge devices that feed the IoT. Edge devices aren't the sensors; they're the widgets that will do something with all the data that sensors provide. We tend to think of the IoT as a vast collector of data that's centrally processed for one purpose or another. The National Oceanic and Atmospheric Administration, for example, collects all kinds of data from a wide array of devices and uses centralized computers to crank out weather forecasts. Wall Street's high-frequency trading apps collect market data and centrally make decisions on financial trades. These are typical big data applications.
But as the IoT approaches its promise of billions and billions of interconnected devices, the big data analysis model fails. And not just because there's too much data to process in a timely way -- though that's clearly an issue. It's simple physics. Petabytes of data just don't move quickly; pesky confines like the speed of light see to that.
Thus, to be effective, edge devices need a level of autonomy. Think of the Google driverless car. Sure, central systems know where the car is and what the local speed limits are, but it's the local sensor array that must tell if the road is bad or if a person or object is moving in the car's path. Local software needs to autonomously take in all this data, including the schedule of where the car is supposed to be and when, and figure an appropriate speed and other actions. This isn't the sort of thing you want centrally calculated and sent over unreliable wireless connections; the car has to be autonomous enough to truly drive itself. Similar technologies have already started to become available in high-end new cars as advanced collision-avoidance systems.
Hand in hand with autonomy is the matter of trust. Your car or your home may have thousands of sensors, and just like any electronics, they'll have limited lifespans. If an outdoor sensor suddenly registers a 20-degree drop in temperature, was it a malfunction or a cold snap? Should your home heating system act on that data or ignore it? If a substation gets a signal to go offline, is it terrorists trying to shut down a power grid, or is it a controlled brownout because of a problem with a generating station, or even more likely, could it have been human error?
Autonomous systems will check and recheck conditions before they take what are known to be extreme actions. Any action will require some redundant analysis, but the further outside the norm, the more data the autonomous system will want to know before it takes action.
The Mars rovers are extreme examples of highly instrumented autonomous systems. With minutes passing between commands from Earth and responses from Mars, a normal control system would be inherently unstable and unusable. The rovers, therefore, have the ability to make their own decisions, and because of that, they've operated exceedingly well -- far beyond NASA's anticipated use period. If the IoT is to have a many-trillion-dollar effect on the economy, it'll come in large part from systems like these and from new businesses not possible without the IoT. Take Zipcar. By embedding sensors and reworking the ignition system in the car and communicating conditions back to headquarters, the company has done away with the rental counter and allowed customers to rent for much shorter periods of time -- less than $10 will get you an hour in a car. Customers love the service, which wouldn't be possible without the IoT. The entry point to business on the IoT isn't low. It takes serious engineering and investment. But the payoff is an untapped market. And as entrepreneurs develop these new business opportunities, the providers underpinning the wired and wireless Internet will have to respond.
iOT vs. M2M
The Blue Jean Effect
Just as in any good gold rush, it's the outfitters (technology suppliers) that are poised to make a killing, especially at first, as IoT applications test their footing. Do a Google search (heck, you can even Bing it) for "Internet of things" or "machine to machine." It's not a pile of clever startups that bubble to the top. It's Verizon, AT&T, Cisco, Intel, SAP, IBM -- the pickaxes and sieves of this gold rush will be technology, data processing and connectivity.
As for IT teams looking to cash in, if you're looking for a lingua franca of the IoT, you're thinking way too narrowly. Remember, this is like cloud computing, but even less well defined. Did the NIST cloud definitions help move functionality forward? Not much, and standards won't help much here, either. There will be specs for smart homes and different standards for smart buildings, which will be completely different from specs for medical equipment, which will look nothing like the standards for logistics or manufacturing. The commonality is connectedness and data processing and not much more. That we're giving this movement a name is more an artifact of hype than it is of anything useful in an engineering or business sense. The Web is evolving to its next stage of usefulness, and that will mean many different, loosely related things.
Think big and wide; the interconnectedness and trackability of everything, along with the capacity to analyze it, all will give rise to huge changes in the way your organization's employees and customers live and do business. But the challenges in making such an expansive vision a reality aren't the least bit trivial. They push the boundaries of technology and require significant new investments in the public and private sectors. The next Amazon or Google may well rise from the IoT -- some smart startup that bets big and wins bigger. But for most, the smart money starts with smaller solvable problems and goes from there.
You May Also Like