Boffins challenge Internet of Things' lust for power

20 billion sensors are all very well but nobody wants to pay the electricity bill


The Internet of Things is going to vastly expand demand for the world's connectivity, and with it, there'll be a corresponding increase in the electricity devoted to computing and communications. The University of Melbourne's Centre for Energy Efficient Telecommunications is kicking off a new project to work out how that impact can be minimised.

CEET director Dr Kerry Hinton explained to Vulture South that just as the tech sector was starting to get a good understanding of how to better manage the power consumption of the global Internet, the Internet of Things revises the architectures in sometimes-unexpected ways.

Take, for example, the assumption that there will be billions or tens of billions of “things” in households, watching the weather, taking care of the lights, managing the refrigerator and other appliances, and all the rest.

It doesn't take an awful lot of these things to reverse our assumptions about traffic: instead of an asymmetry in which most traffic is downstream, the IoT starts to assume either symmetry or even more traffic flowing upstream.

Why would that be? – Citing the well-known predictions from the likes of Cisco and Ericsson that there will be billions of things, “that data has to be processed somewhere,” Dr Hinton said. So there will be gateways that “aggregate traffic from the monitors, and if needed, send that off to the data centre”.

How does that become a power efficiency challenge? – because today's home Internet gateways most commonly send a much smaller amount of upstream data than they receive, and as a result, use less powerful transmitters than the ones in the DSLAM devoted to streaming video in your direction.

That's just one example of the kind of question the CEET is going to work on in its new energy-efficient Internet of Things studies.

Wireless communication, Dr Hinton told Vulture South, also involves some tradeoffs that the CEET project will study. For example, the very low power wireless in “things” needed to keep their battery demands low means that the gateways receive will be of a very poor signal-to-noise ratio ratio.

“For low-signal-to-noise wireless technologies, the trade-off is how much energy goes into processing in the gateway versus how much energy is available in the 'thing',” he explained.

And there's the question of where data needs to be processed. Dr Hinton pointed out that the world is learning a lot about how cloud computing can improve energy efficiency, but that in the IoT world we're yet to understand the trade-offs that will be needed.

Putting too much of the processing in the gateways is inefficient, he said, so the question is “how do you shift some of those functions into the parts of the network where it's most energy-efficient?”

If there are too few data centres, all that will happen is that “you're going to clog the network with traffic from tens of billions of things.”

So the centre will look at modelling to find the “sweet spot” to work out how decentralised Internet of Things processing must be while still reaping the energy-saving benefits of cloud computing.

The work could also feed into understanding how to design Internet of Things devices and applications where the electricity grid doesn't reach – remote areas or in the developing world – so that 'things' can be solar powered without their batteries dying within an hour of sunset.

CEET partner Bell Labs says the work will feed into that organisation's sustainability research agenda. ®


Keep Reading

Biting the hand that feeds IT © 1998–2021