Skip to main content

PUE - What future for power usage effectiveness ratings?

This article was originally published on Technology.Info.
As part of our continuing strategy for growth, ITProPortal has joined forces with Technology.Info to help us bring you the very best coverage we possibly can.

Are efforts to improve the energy efficiency of data centres starting to hit a wall? Survey results recently released by Uptime Institute, a market research and industry consulting firm, would seem to suggest that that’s the case.

Speaking at the company’s annual Symposium in Santa Clara in May, content and publications director Matt Stansbury had the following message for attendees: “We’re not really getting anywhere.”

The vast majority of the data centre operators who participated in Uptime Institute’s annual survey measure efficiency using the Power Usage Effectiveness (rating), which compares the total amount of power that enters a data centre facility to the amount of power used by the IT equipment in that facility. This gives an indication, meanwhile, of the energy consumption of ‘overhead’ - the lights, cooling equipment, air conditioning and other equipment not directly active in the processing and storing of data.

Ideally, a data centre’s PUE rating should be as close to 1.0 as possible - and for four years, the Uptime Institute has measured the average efficiency ratio across its survey base. This year, respondents reported a PUE ratio of 1.7. That’s up from 1.67 in 2013, although still down on the 2012 and 2011 averages of 1.8 and 1.89, respectively. But, in other words, the downward trend stopped some time last year.

Is PUE as a measure of data centre energy efficiency flawed?

But the situation may be more complicated than that. Many in the industry say that the PUE rating, as a measure of data centre energy efficiency, is flawed. It doesn’t give a true picture, say its detractors, without accompanying information about how it was measured. The simplicity of the rating makes it too easy for data centre operators to dissemble, they say.
While some PUE supporters, meanwhile, concede it has shortcomings, most agree that it’s the best measure they’ve got to work with. It’s simple, and workable. More importantly, it’s already a de facto industry standard and on course to be approved by international standards body, the ISO.

What alternatives to PUE are being explored?

Over the years, this argument has rumbled on, with other approaches proposed from time to time. The most recent, the DCEM (Data Centre Energy Management) Global KPI, comes from ETSI (the European Telecommunications Standards Institute).
“The DCEM Global KPI offers a reliable operational measurement method and includes the energy reused for other purposes than data centres,” says Dominique Roche, chairman of two committees within ETSI, in a June 2014 press release . “We also took into account the Kyoto protocol on the reduction of greenhouse gas emissions. Data centres can therefore be part of an energy-efficient, sustainable world.”
While few in the industry would argue that this is an admirable goal, it’s far from certain at this point that DCEM will be able to gain any real traction. As a recent (and enlightening) analysis [] by Peter Judge in makes clear, it is enormously complicated, classifying data centres into four sizes, allowing many different modifications to the score for renewable energy use and waste energy re-use, and then grading data centres in each of the four size-groups on a scale of A to I.
Having analysed the new measurement, Peter concludes: “There is a danger that [the European Union] may pick up an over-complex standard. This could create perverse incentives and unforeseen consequences, and could produce a wasteful parasitic industry, making and marketing DCEM measurements simply to achieve certification.”

Power wastage from redundant IT equipment driving up PUE will continue to watch this issue with interest and report back - but in the meantime, Uptime Institute is working to raise awareness of an issue that continues to scupper energy-efficiency efforts, regardless of how their outputs are measured. That’s the issue of so-called ‘comatose’ servers; machines that have been abandoned by application owners but are still racked and running, “squandering power at the plug, but also wasting data centre power and capacity.”
“Comatose servers are like weeds in a garden - rampant, insidious - and abatement is an ongoing responsibility,” says Uptime Institute’s Stansberry. That’s why, he says, the firm encourages companies worldwide to participate in its Server Roundup, an initiative to eliminate redundant IT equipment.
At this year’s Symposium, it presented financial services giant Barclays with an award for its work in this area: it removed 9,124 physical servers from its IT infrastructure in 2013 alone.
And the benefits are already clear, according to Paul Nally, a Barclays director. “We are seeing reductions in power, cooling, rack space and network port utilisation - all of this, while our usable compute footprint goes up, giving us room to continue to grow the business,” he said at the award presentation. “Servers eliminated in 2013 directly consumed an estimated 2.5 megawatts of power. Left on the wire, our power bill would be approximately $4.5 million higher than it is today.”
Other data centre managers might do well to follow Barclays’ lead. According to Uptime Institute’s survey, around half of respondents don’t have scheduled auditing for comatose servers. Half of these, meanwhile, attribute this to a lack of management support, one-third say their resources are too limited and 20 percent said the return on investment did not justify the effort.
Read on to find out how aircraft manufacturer Airbus is taking a more ‘modular’ approach to architecture in the area of high-performance computing (HPC).