Can you measure your dirty data?

It’s hard to ignore energy use in aviation: after all, we see planes fly overhead every day, hear their roaring engines and see evidence of their fossil fuel-powered emissions through the contrails they leave behind in the sky.

With the internet, though, it’s easy to forget that every website, every Google search, every email and every Tweet has an energy (and carbon) footprint. We don’t have to make any special efforts to fuel our laptops — no trips to a filling station required — and there’s not a hint of emissions as we blissfully surf through cyberspace.

Of course computers use energy. We know that. But it’s something of an inconvenient truth that computing, measured on a global scale, is almost as much of an energy hog as aviation. While air travel accounts for more than 2.5 per cent of the world’s total energy demand, data centres aren’t far behind with a consumption level of between 1.5 and 2 per cent. Both sectors also appear to be in a race to see which can grow faster: domestic air travel is set to double over the next 20 years, while data centre energy demand is expanding by 12 per cent a year.

Data centre operators who want to measure how efficiently they’re running generally rely on a metric called power usage effectiveness, or PUE, which is calculated by dividing the total amount of energy used by a data centre by the amount of energy used by its IT infrastructure alone. In an ideal data centre, the energy used for IT would be the only energy necessary (that is, the net power consumption for lighting, cooling and other functions would be zero), yielding a PUE of 1.0.

The trouble is, PUE measurements aren’t always consistent from one data centre to the next, making it hard to compare facilities fairly. A task force at the Green Grid Association has been working to fix that problem, though, and has just released new recommendations for measuring and reporting energy efficiency in data centres.

The task force has three specific recommendations:

• Make PUE based on source energy consumption the preferred energy efficiency metric for data centres;
• Data centres should, at a minimum, measure their energy consumption at the source: the uninterruptible power supply, or UPS. Over time, however, the task force expects that it will become easier to measure energy consumption directly for each component in the data centre, that is, servers, storage, networks, etc.
• Dedicated data centres should count all the energy provided by its utility company for its PUE measurement. Mixed-use data centres, on the other hand, should base their PUE measurements on the total energy needed to operate the facility (including cooling, lighting and support infrastructure).

The Green Grid Association task force also specifies that the PUE should be calculated the same way for data centres powered by renewables as for those connected to the grid.

“The rationale for this is that the purpose of the PUE is to evaluate how efficiently energy is used in the data centre, regardless of the energy supply,” the recommendations state. “PUE is not a metric to reward or penalise the efficiencies of how electricity is brought to the data centre, it is a metric to identify how efficient the electricity is used from the data centre boundary to the IT (e)quipment.”

PUE calculations also shouldn’t change if a data centre reuses waste energy, the task force says.

“If the energy is reused within the data centre, the value of the reused energy will already be contained within the PUE calculation and needs no special consideration,” the recommendations state. “While the effort to conserve energy is laudable, it cannot be accounted for in the calculation of the data center PUE due to the strict definition of PUE used in these recommendations. Currently there are on-going industry efforts to define a metric that could be used to account for this beneficial use, but it is specifically excluded from PUE.”

The result is that, while a PUE measurement can give us a good idea of how efficiently a data centre uses the energy it consumes, it offers absolutely no insights into how much carbon emissions a data centre might be responsible for or how much greener one operator is than another. A data centre that gets all its electricity from coal-fired power (Apple’s, Facebook’s and IBM’s reliance on coal power, for example, have earned them criticism from Greenpeace) could still have a sterling PUE of 1.1, while a facility that’s completely solar-powered could earn a less desirable PUE of 1.3 or higher.

“Metrics like PUE do have value in helping data centre operators benchmark the design and efficiency of their facilities by providing
an objective metric that drives efforts to improve facility efficiency,” Greenpeace noted earlier this year in its report, “How dirty is
your data? A look at the energy choices that power cloud computing.” “However, many companies are affirmatively using low PUE ratings to communicate externally that their data centres are ‘green’ and sustainable without accounting for the full environmental picture.”

The Green Grid Association recently announced new metrics to better reflect a data centre’s sustainability, including carbon usage effective (CUE) and water usage effectiveness (WUE). A single measurement that could give an overall picture of how green and efficient a data centre is, though, would be even more useful for consumers.