“Networked” and “wired” used to be synonyms. Downstream devices were hardwired to upstream controllers or PCs that contained the bulk of the network’s intelligence. Upstream devices continually polled the “dumb” downstream devices for data, even if the data wasn’t changing very often.
If “dumb” downstream devices lost their network connections, their data was lost. To make the model work smoothly, one of a network designer’s most important goals was to get as close as possible to what the telcos famously call “Five Nines” (99.999) uptime. And the closer you got to perfection, of course, the harder it was to achieve the next incremental improvement.
While wired networking is far from obsolete, advances in wireless technologies like Wi-Fi, Bluetooth 4.0, and cellular data communications are opening the door to entirely new network functions and topologies. It’s becoming possible to wirelessly network-enable just about any device, just about anywhere—even legacy serial devices. But if wireless connections are going to be used to their full potential, they can’t be hobbled by the old network model.
Power Management Gains Importance
A wired network design normally assumes easy access to power. Wireless networks can’t do that. As they continue to expand the network’s edge, they’re not only including locations that are beyond the reach of cable connections, they’re quite literally taking data communications right off the power grid. More and more remote wireless devices and sensors are being tasked to function with local power only.
Suddenly, power budgets are a huge factor in network design. Network designers and engineers don’t want to make recommendations for one-month battery replacement cycles or for batteries that are so large they render a project impractical.
So, there’s a need for network designs that use power efficiently enough that the remote network nodes and devices won’t go dark. Traditional “always on” communications schemes won’t do the job. They would immediately consume whatever power budget was available.
Fixed sampling rates, for example, dictate a client/server type architecture in which a “dumb” network node must be in active listening mode most the time. Fixed sampling rates also require the wireless node to respond with data at every specified interval, even if the data is of no value. That’s not an issue for a wired network node, but remote wireless nodes often need to avoid making unnecessary charges on the power budget.
Five Nines Of Uptime
Today’s “smart” devices don’t have to obey the old rules. ICs are becoming steadily smaller and more powerful, which means local intelligence can be installed all across the network. More and more devices will be equipped with situational awareness, employing parameters like power, network availability, and the status of surrounding nodes to make independent decisions about their own operations.
Rather than remaining in constant communication with the network and operating at full power at all times, “smart” devices will collect data, time stamp it, log it, and report the data whenever a network connection becomes available. Even if the network is down for minutes or hours, this “store and forward” model ensures that data isn’t lost.
Smart devices are also acquiring the ability to report on exception. Again, instead of attempting to remain in constant communication, the node only reports when its internal logic tells it that data has gone out of defined boundaries. It doesn’t need to report until it has something meaningful to say.
Designing for “often available” connections rather than “continuously available” connections not only makes Five Nines uptime unnecessary, it conserves energy that would otherwise be wasted on idle chatter. By staying abreast of these advances in wireless technology, network engineers will be able to specify “smart” devices to help them expand network capabilities while simultaneously managing their power budgets.
Making connected devices more independent doesn’t relieve the central controller (or master) of the need to know whether a node is nonfunctional or merely being quiet. (Let’s say that a water tank is half full, has been that way for days, and the remote sensor has nothing of interest to report.)
In the old network model the nodes had to remain powered up, eating away at the power budget the entire time. But a “smart” node can be given a simple “heartbeat” function instead. By pinging the master every now and then, it lets the master know that while it may be one very dull node, it isn’t a dead one.
This isn’t just pie in the sky. The University of Michigan has already developed a low-power, smart sensor system that uses most of these techniques. At just 9 mm3, it’s smaller than the eraser on your pencil, but it’s solar powered, has an internal battery, and is equipped with its own processor (see the figure).
Called the Phoenix, the processor employs a unique power gating architecture and an extreme sleep mode to achieve ultra-low power consumption (http://www.engin.umich.edu/newscenter/feature/smallsensor/). A more recent version of the system added a compact radio that needs no tuning to find the right frequency, enabling node-to-node communication in wireless sensor networks (http://ns.umich.edu/new/releases/8278).
“Smart” network nodes will only continue to get smarter, and additional advances in wireless technology, power sourcing, and power harvesting will make them increasingly independent. But there’s no such thing as a free lunch. While it may be true that network designers are gradually being freed from the need to pursue Five Nines uptime, it’s only so they can spend their time perfecting power budgets and making incremental improvements in wireless security.