Thursday, 15 November 2012

ICT energy does not obey Moore's Law


The processing power of ICT hardware has followed Moore's Law for several decades: energy efficiency has improved nowhere near as rapidly, and something has to be done.
Late last week the Centre for Energy-Efficient Telecommunications (CEET) at Melbourne University and Alcatel-Lucent won the Best Research & Development Collaboration award at the Business and Higher Education Round Table.

It was the second award for their collaboration. Earlier this year Alcatel-Lucent won the telecoms industry's ACOMM award for environmental responsibility for its work with CEET. The company's response to that, however, was not one of unalloyed gratitude: it was described by Alcatel-Lucent Australia managing director, Seán O'Halloran, as "a hollow victory."

The reason? There were no other contenders for the award, a fact that O'Halloran says reflects badly on the priority the industry as a whole is assigning to its surging energy consumption.

Announcing receipt of the latest award, CEET said: "Today communications technology consumes up to two percent of the world's electricity supply but…rapid growth in broadband demand could lift Internet energy consumption to five to 10 percent of the world's electricity supply and associated emissions by the year 2020."

Scary stuff. Or scaremongering? Most likely the former. Announcement of the collaboration award co-incided with publication of the latest edition (Vol 62, No 5 November 2012) of the Telecommunications Journal of Australia (www.tja.org.au). In addition to a paper by O'Halloran - the source of his 'hollow victory' comment - it contains several others focussed on the nexus between information processing and energy consumption.

Perhaps the most striking observation to emerge is that there is little relationship between the impressive advances that flow from the fulfilment of the prediction best known as Moore's Law and increasing energy efficiencies.

In another paper in the journal CEET's deputy director, Kerry Hinton, said that if a unit of processing power cost $3000 in 2000, the same money in 2015 would likely buy 64 units, but while it might have cost only $150 per year to power and cool that one unit in 2000, the cost to power and cool those 64 units in 2015 would likely be $3000 per year. In other words, only a threefold efficiency increase, assuming constant power costs. (The cost to power and cool 64 units at 2000 rates would be $9600.

This changing metric of power consumption per dollar of processing power capex produces some startling statistics - spelt out in another TJA paper by Bob Hayward, CTO for CSC Australia and Asia.

- Globally, data centres consumer more power than the entire country of Russia;
- the largest user of electricity in the UK is Government ICT;
- electricity consumption by data centres is doubling ever seven to 10 years.

And, to make it personal: an Australian who regularly visits the virtual world (frequency and type of visits were not specified) consumes as much electricity in a year by doing so as the average citizen in Brazil for all purposes.

Taken together there is only on conclusion from these metrics: the present trajectory is not sustainable.

Solutions posited in the various TJA papers covered a broad range. There is significant effort being directed through CEET and a global organisation known as GreenTouch (www.greentouch.org) to greatly increase the energy efficiency of information processing technologies. GreenTouch has the goal of reducing the energy consumption of telecommunications networks by a factor of 1000.

That may seem inordinately ambitious, but if you marry it to the projection from Nokia that mobile network capacity will have to increase 1000 fold by 2020 then it amounts to merely maintaining current levels, in one area at least.

Another argument is that ICT can do much to increase energy efficiencies in other areas, thus compensating for its gargantuan energy appetite.

No doubt progress will be made on these and other fronts, but it's seems clear that substantial, and rapid progress will be needed on several fronts if current projections on the rise of cloud computing, number of connected devices, global internet traffic and any other indicator of ICT momentum you care to name are to be achieved.

O'Halloran sees the opportunity for Australia to become a world leader in the field. He expects momentum to build quickly in areas like consumer awareness and the development of a framework for an Internet energy star rating, which he describes as "an extraordinarily complex initiative already on the drawing boards at CEET."

But he says that progress will not be achieved without the broader Australian telecommunications sector "having the courage and vision to encourage collaboration between some unlikely players."

Given that many industry players are in fierce competition with each other this seems like a call for some organisation that can transcend that competition to take some initiative to seize this significant opportunity.

This article first appeared on iTWire, Australia's leading independent IT&T news and information source.

No comments:

Post a Comment