There is a lot of discussion about energy efficiency in data centres – but is it aiming at the wrong target?
As data centres have increased in size, so has the size of their energy bill, and the amount of emissions for which they are responsible. This has led to campaigns from environmental groups like Greenpeace for data centres to cut their energy use, while the industry itself, through groups like the Green Grid, has concentrated on making data centres more efficient, so they can do their computing with less energy.
This week, a Forrester report has suggested that the industry’s efforts may be missing a trick. While concentrating on the efficiency of data centres, it is not majoring on the extent to which they are using renewable energy, or on the progress towards dealing with e-waste.
This is slightly unfair. The Green Grid has proposed CUE (carbon usage effectiveness), a ratio based on its its PUE (power usage effectiveness) measure of energy efficiency, which lets data centre owners measure and quote the carbon intensity of their data centre.
And e-waste is a real problem, but one which was always going to be a tough one, since it involves actual long term costs. While efficiency savings often reduce expenditure on energy, disposing of hardware responsibly is always likely to cost more than dumping it. Regulations will have to force action here – and parts of the industry will oppose or delay these measures.
Greenpeace is, of course, pushing both of these issue – and its attitude contrasted with that of the data centre efficiency movement last week, when Greenpeace met the Green Grid on stage at the 451 group’s Hosting and Cloud Transformation Summit in London.
The Green Grid, and Dr Ian Bitterlin of Ark Continuity, both argued that data centres should be the good guys in Greenpeace’s eyes, as their use of IT is more efficient than the thousands upon thousands of under-utilised servers in racks and rooms round the country.
Bitterlin suggested that, by reducing the average PUE of servers from 3 down to 1.2, the cloud could cut the power used by data centres by more than 60 percent.
In slides he shared with eWEEK Europe, he went further, suggesting that environmental campaigners should shift their attention from service providers to the hardware makers – since servers waste power even when they are idle.
Hardware makers have already done a great job, he said. In 1997, the world’s fastest supercomputer at Sandia national Laboratories in the US delivered 1.8 teraflops, using 800kW and 150 square metres of raised floor space. By 2006, the same power could be delivered by a Sony Playstation 3, using less than 0.2kW, and 0.08 square metres of space.
But despite this, their record is poor on idle power. “Even Energy-Star rated servers in 2011 consume as much as 65 percent of their peak power when doing precisely zero IT work,” said Bitterlin.
Bitterlin wants to see the graph of power cosumption become more linear, so low utilisation means less power used – moving from the black line in the graph to the red one.
We’d like to see that too, as well as movement on all the other fronts of the sustainability battle.
Targetting AWS, Microsoft? British competition regulator soon to announce “behavioural” remedies for cloud sector
Move to Elon Musk rival. Former senior executive at X joins Sam Altman's venture formerly…
Bitcoin price rises towards $100,000, amid investor optimism of friendlier US regulatory landscape under Donald…
Judge Kaplan praises former FTX CTO Gary Wang for his co-operation against Sam Bankman-Fried during…
Explore the future of work with the Silicon In Focus Podcast. Discover how AI is…
Executive hits out at the DoJ's “staggering proposal” to force Google to sell off its…
View Comments
This again highlights the fact that data centre providers are hamstrung by the server and other IT manufacturers. We can squeeze PUE to death but we reach a brick wall against server power usage.
PUE does not address the amount of energy used by the data centre but only the portion of energy used by IT.
So, true energy savings can only happen once the inefficiencies in server design are addressed.
That is where real results will be achieved.