Graphic from May 1 2002 issue ofCIO.

The Device Internet

Fred Hapgood

The good news for CIOs is that there is no mystery about the next big step in IT. Howevermuch fog there might be about biotechnology or robotics or the future of dirigibles virtually no one would disagree that over the next few years the business of this trade is going to lie in connecting hordes of so-called "dumb devices" to the network. The bad news is that in the short to mid term this task may very well overwhelm the resources of the average IT office and in the long term, for better or worse, change it out of all recognition.

This future is obvious because the potential is enormous. There are two classes of dumb devices: sensors (such as cameras, microphones, counters, meters, scanners, scales, location or position sensors), which measure physical changes, and actuators (motors, lights, switches, valves, manipulators, displays, loudspeakers, printers, locks), which impose them. Many devices, such as telephones, have both.

The primary benefit of wired devices is that they communicate or connect with just the right person at the right moment. Instead of maintenance technicians visiting each machine on a set schedule, or checking every machine during a down period, networked sensors can signal exactly when and where maintenance is needed to the best (perhaps the nearest, or best equipped) technician for that job, insuring that every visit has a payoff. A networked security system, in which the relationship between passcards and specific locks can be defined anywhere, makes it easy to assign the right bundle of security functions to each management level. When David Graham, Director of IT Operations at Vignette, a content services provider in Austin, Texas, manager, installed network monitoring devices that were themselves networked (from Netbotz), he was able to patch Facilities Management right into the heat sensors. "It took a bit of the pressure off of us," he says.

Recently the University of Louisville networked their printers (using digital printers from Xerox), allowing users to drive print jobs to the most efficient and convenient printer for each job. According to Ronald Moore, Campus Technology Chair at the University, the returns were immediate. "It's nice to see an IT solution that actually saves money," he muses. His experience is not unique: Harbor Research, a Boston strategic consultancy, reports that device networking initiatives typically pay for themselves within 24 months or less.

Device networking also leads to more flexible services, since the control interfaces associated with networks are usually more powerful than those packaged with stand- alone models. Recently Don McGill, Director of Voice Engineering for CNET Networks, the SF technology news company, moved the company's phones onto their network in order to acquire the features (like caller history management) that come with telephony in desktop environments. However, he cautions that it can be frustratingly difficult to get people to understand that they really do need training to use networked devices. "If you announce a class in Java Programming," he says, "people understand they need it. A class in using the phone tends to get skipped." (Moore of Louisville made the same observation about training for networked printers.)

Finally, up till now most devices have been monitored or controlled by a single type of user; i.e., security cameras were watched by security personnel. Connecting devices to the net give every constituency in the enterprise access, allowing a host of new applications to be built out of the same peripherals. In the case of security cameras, marketing might use the feed to see which product displays attract traffic; personnel, to select candidates for extra training; customer relations, to let clients watch their particular job be executed, etc. Facilities management might control room capacities by setting totals on how often doors can be locked and unlocked.

These attractions are why research companies like Harbor predict that in five years device connectivity will be consuming nine billion chips a year, accelerating to more than 40% of "relevant electronic and electromechanical devices" by 2010. Forrester Research reaches similar conclusions, predicting sales of 14 billion network enabling chips by that year. In short, in a few years networks will start to be primarily about communication among devices, not among people.

This change probably implies a radical transformation in the character, structure, economics, and management of networks. Bill Peisel, CTO of NetSilicon, a device connectivity vendor in Waltham MA, suspects many of these devices will be communicating directly amongst themselves, making peer- to- peer a more common communications model than C/S. Security will be more about good authentication than good encryption. "It's not that important to encrypt the room temperature;" Peisel says. "It is important to deny access to the 13 year old who thinks it would be really amusing to turn off the heat in the middle of the winter." Device-oriented networks will probably need to be more robustly provisioned, with redundancy and higher levels of automated maintenance built in at every level. "We can lose a patient in seven seconds," says Reza Sharafi, an engineer at Welch Allyn, who makes networked medical sensors. "If the network goes down we lose everything. We don't want to hear that we can 'just reboot'."

Device-oriented networks may well blur the lines that currently define financial responsibility for maintenance, management, and procurement. For instance, Moore of Louisville says that device networks seem to increase the pressure for upgrades. When you expose everyone to all the devices of a given type in the enterprise, users who were once happy with the machine in their office start to lust after the most advanced feature set available. When five different departments suddenly get their hands on a device originally designed for just one user type, it should not be surprising if IT is deluged with wish lists from these new users.

Yet supposing security is perfectly happy with the old cameras (for instance). Who pays? Suppose security only needs to store the video feed for 24 hours, but marketing or human resources want to preserve it for months? Who gets billed for storage? Who will be responsible for supporting the software that will implement whatever bookkeeping formulas emerge from all these complex usage patterns? How will priorities be set in a time when dozens of devices are waiting to be connected, all with a case? When HVAC is a network device, IT issues like security and provisioning will affect the heat and Facilities issues like maintenance and procurement will affect the network. How will that be resolved? Will CIOs have to learn more about field operations or field managers more about networking?

It seems unlikely that the management structures that worked when the ratio of intercommunicating entities on an enterprise backbone to users was two or three to one will be as useful when that ratio is ten or even a hundred times higher. ("We are now at the point where we can network individual gears and bearings," observes Joe Carpenter, CIO of Swantech, a Fort Lauderdale company that makes networked diagnostic equipment for industrial machinery.)

Some observers suspect that field operations might start to take over a number of the more finely grained network issues. Chantal Polsonetti of the ARC Advisory Group, an automation consultancy based in Dedham, Mass., points out that field and network people have different operating cultures. CIOs often come from a store and forward world in which fractions of a second don't matter that much, whereas in the world of physical devices, where functions can have safety implications, even small delays can have large consequences. While the technology could wait for backbone upgrades, Carpenter of Swantech argues that this delay is not necessary, since adding local resources can make devices smart enough to tackle many network issues on their own. Computation allows them to model and then adapt intelligently to network behavior; local storage lets them take on responsibility for issues like reliability (that is, when the network goes down, device storage kicks in).

Another name for the transformation of 'dumb devices' into 'smart devices' is automation. As processes move to higher degrees of automation, the exceptions, the moments when the system really does break done and human intervention is called for, will become more exotic and unprecedented. Fewer hours will be required to keep the enterprise backbone running, but each hour that actually is required will call for much higher levels of skill.

In theory this seems like an ideal environment for the elaboration of a globally outsourced support model, with hundreds of thousands of super-experts, each deeply educated in a few very specialized support issues, monitoring the device networks of clients around the world, and dealing with crises as they erupt. In other words, the migration to device networking might shrink the number of routine IT support issues (through smarter devices and more local management) while outsourcing the remaining, tougher problems.

It would seem like a gloomy prospect for the conventional IT office, except for one consideration: the super- experts are going to have to come from somewhere.