Measuring and reporting data centre PUE


Measuring and reporting data centre PUE

Mark Fontecchio, News Writer and Matt Stansberry, Executive Editor

The power usage effectiveness (PUE) metric allows data centres to measure their energy efficiency and, in some cases, publicise the fact for bragging rights or marketing gain. But the Green Grid, which created the metric, is trying to reclaim the reins on how to measure and report it.

About two years ago, the

Continue Reading This Article

Enjoy this article as well as all of our content, including E-Guides, news, tips and more.

By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.

You also agree that your personal information may be transferred and processed in the United States, and that you have read and agree to the Terms of Use and the Privacy Policy.

Safe Harbor

Green Grid, a nonprofit consortium of vendors and users that focuses on data centre energy efficiency, released the PUE metric. It was simple: Take the amount of total data centre facility energy and divide it by the energy that gets to the IT equipment. The more efficient the data centre, the closer that number is to 1.

But soon it became apparent that this number hinges on several factors, and not publicising those factors along with your PUE only leads to incomplete information.

Why publicise data centre PUE?
When the Green Grid first published the PUE specification, it was intended to be an internal measure, according to board member John Tuccillo, the vice president of global industry and legislative initiatives at American Power Conversion (APC). But as Tuccillo put it, "there is a real value to being able to share data and compare data" across organizations. And some companies did exactly that.

For Advanced Data Centers (ADC), it's about selling its product. Last year, the Sacramento data centre colocation company announced that it was designing a data centre with a PUE of 1.1, which is very low. (The Uptime Institute estimates the average to be about 2.5.) But why did ADC throw that number out there? Simple: competitive edge.

ADC President Michael Cohen said that the company bills customers on multiple components. The first is rent based on square footage and power capacity. The second is the metered power used by IT equipment in a data centre space, measured at the uninterruptible power supply (UPS). The final is a percentage of the cost it takes to cool the building.

It's that last component, which Cohen called the cooling factor, that can make a difference. The more efficient the data centre cooling infrastructure is, the lower the PUE. By advertising a low PUE, ADC has sent a message to potential clients that it will be cheaper to colocate equipment there.

"A lot of our customers are saying they want the benefit of that top-tier engineering," Cohen said. "If they can say they're green and can save money doing it, that's the best place to be."

For other companies, publishing PUE may come down to bragging rights, or fulfilling a public promise to go green. Google, for example, publishes its PUE as a way to show the public that it does "no evil." Google's green data centre site aims to show that the company is energy and environmentally conscious.

But there's a problem: A PUE can depend on a lot more than just power, and that's why many believe the number should be only internal.

"There has been a lot of debate on whether folks should [make PUE public]," said Roger Tipley, a Green Grid board member and engineering strategist at Hewlett-Packard. "It detracts from the incentive of doing continuous improvement. The way you measure your data centre might be different from how the guy measures it across the street. It's apples to oranges. The point is to measure the overall impact of the infrastructure, not try to game the metrics."

Factors that affect data centre PUE
ADC is building a data centre in Sacramento, which has a temperate climate that can support air-side economizing (a method that uses outside air to cool a data centre). ADC hopes to use air-side economizing 75% of the time. The benefit is that there's no need to power mechanical chillers to provide cool air.

Needless to say, it's not feasible to run air-side economizers 75% of the time in Phoenix, Houston, or most places in the country. The climate doesn't support it. And because of that, a Sacramento data centre's PUE will necessarily be lower.

Another factor is data centre power infrastructure. How redundant is it? A company building redundancy with a 2N UPS configuration (twice the UPS than the required load), along with backup generators and chillers, could have a higher PUE than a company without redundancy.

Yet another factor is where exactly you measure power to the IT equipment. At the UPS? At the power distribution units (PDUs)? Or at the servers and switches themselves? This matters, because as power travels from the UPS to the PDU to the IT equipment, there are power losses, however miniscule. If you measure at the UPS, your PUE will be higher than if you measure at the equipment level.

When you measure PUE is another key ingredient. A PUE will be lower if the servers are at full load, because they consume more power at that point compared to the data centre's total power footprint. So if you take one-time measurements at ideal times, you can game the system.

How to report PUE publicly
According to the Green Grid, all these factors should be part of any publicly reported PUE.

"We know people will start using it for comparison," said Mark Monroe, a board member on the Green Grid and sustainable computing director for Sun Microsystems. "We want to detail how to measure and report it to see that it's verifiable and reportable."

The Green Grid started that process by publishing a research paper on reporting PUE/DCIE. The paper defines three levels of reporting the figure depending on where and when measurements are taken:

  • Basic. IT equipment power is measured at the UPS, facility power is measured at a data centre input power source, and measurements should be made at least once a month.
  • Intermediate. IT equipment power at the PDU, facility power is the data centre input power minus the shared cooling, and measurements are taken daily.
  • Advanced. IT equipment power at the equipment level, facility power is the data centre input less shared cooling, plus power for building lighting and security, and measurements are taken continuously (every 15 minutes, for example).

Tipley recommends measuring PUE every 15 minutes.

"A typical data centre's PUE score drifts quite a bit," he said. "IT loads go up and down, and total energy consumption has a scalability factor. How variable is your data centre energy consumption? Is it a variable IT load or infrastructure?'

The Green Grid is also working on paper that defines segments of the data centre industry. Factors include whether the site is primarily for testing or production; if the business is financial, health care, data centre colocation or something else; and what the data centre's tier level is. That way, like data centres can compare PUEs.

Let us know what you think about the story; email Mark Fontecchio, News Writer.