Energy Star for data centers has bright sides, dark sides

Though overly broad, rating system will encourage measurement and efficiency

You've no doubt seen the familiar Energy Star label emblazoned on computers, refrigerators, televisions, and other electronics and appliances, indicating that the machine you're using is among the crème de la crème of energy efficiency. In the not-too-distant future, a select group of organizations will be able to proudly display a similar seal on their data center walls. Whether the seal accurately or fairly reflects a high level of energy efficiency, however, remains to be seen.

After many months of crunching data center statistics and working with players from all walks of data center life, the EPA has announced that it will release on June 7 the finalized Energy Star performance rating for data centers in Portfolio Manager, the group's energy management tool for tracking and assessing buildings' energy and water consumption.

[ New tools from Viridity aim to boost data center efficiency. | Keep up on green IT trends with InfoWorld's Sustainable IT blog and Green Tech newsletter. ]

The new rating is based heavily on the popular PUE metric, which compares the amount of energy a data center uses as a whole -- for IT, power delivery, cooling, and so on -- to how much it uses solely for powering IT gear. The lower a PUE rating, the more efficient the facility. According to the EPA, PUE generally ranges from a 1.25 (good) to a 3.0 (not so good).

Rather than being a raw PUE calculation, though, the new Energy Star rating -- ranging in scale from 1 to 100 -- is based on how the data center's actual PUE compares to its predicted PUE. That prediction is calculated based on the facility's annual IT energy use and essentially tells you what the average PUE would be for data centers similar to yours in size and operating characteristics and with the same annual IT energy use.

Achieving the predicted PUE would earn you a rating of 50, meaning your facility has average efficiency for its size, energy use, and IT load. A facility with a score of 75 or higher -- which means it's among the top 25 percent of energy-efficient data centers -- is eligible for the Energy Star.

The model, as you would expect, is intended to reward companies that embrace more energy-efficient practices. In a presentation to The Green Grid earlier this year [PDF], the EPA provided an example of how using an economizer, for example, would boost a data center's rating. In the scenario, two identical data centers -- same UPS energy consumption, same size, same climate -- would have the same predicted PUE of 1.87.

The differentiator: One data center would employ an economizer for cooling, which would result in 20,000 fewer MBtu consumed per year. That, in turn, would reduce the facility's actual PUE from 1.73 to 1.64. As a result, its Energy Star rating would be 70 instead of 60.

Two questions in particular spring to mind when considering this data center rating: Is it effective, and is it fair?

The system is certainly effective insofar as it should motivate data center operators to measure the energy efficiency of their facilities, the first critical step in reducing waste. Additionally, companies that boast of exceptionally low PUEs and high levels of efficiency can now prove it in a standardized, verifiable way.

One of the drawbacks, however, is that it's based heavily on PUE, which has garnered some criticism for not being an entirely effective tool for assessing a data center's overall efficiency. One of the most common cases against PUE: A data center has 100 servers working at 80 percent utilization, 24/7. Its actual PUE is 1.25. An absolutely identical data center has the same 100 servers -- but they're idling 24/7, while drawing the same amount of energy. The second facility's actual PUE is also 1.25, despite the gross inefficiency. In other words, a data center bearing the Energy Star label may still be inefficient.

[ The Uptime Institute's Kenneth Brill cautioned not to get too caught up in PUE.]

Jon Haas, technical committee vice chairman for The Green Grid, said that the group supports the EPA's efforts and its decision to embrace PUE as central to the rating. However, Haas, who is also director in the Eco-Technologies Programs Office at Intel, indicated that the rating is but a first step in developing a meaningful energy-efficiency standard for data centers. "We're starting to talk about [IT hardware] productivity and efficiency. We have efforts in that space to realize those metrics. That will take some time; it requires a lot more heavy lifting," he said.

In terms of the fairness of the rating, I see potential drawbacks. Among them, the rating is designed to be one size fits all; you won't find different ratings for different categories of data centers. Consider two virtually identical data centers: same size, same location, same amount and type of IT equipment, but one is Tier I and the other is Tier III or IV. A higher-tier data center requires a higher level of redundancy to ensure uptime, which necessitates more backup equipment.

That, in turn, means a higher PUE compared to the Tier I facility, out of necessity, not because of inefficient practices. Yet the Energy Star rating puts the Tier I data center at a disadvantage because it applies the same predicted PUE to both facilities. It's akin to denying a workstation an Energy Star rating because it's not as energy-efficient as a notebook, even if it's more energy efficient than most other workstations.

The EPA says it did consider tier level when developing its model and found that "tier level did not show strong, statistically significant correlations with energy consumption."

Additionally, respondents within the industry requested that tier level not be factored in for various reasons. Among them, some data centers have multiple tiers, and "normalizing for tier level provides a disincentive for efficient design."

Haas, on the other hand, said he would expect some kind of segmentation down the road to recognize that data centers fall into different categories. The EPA does have that sort of segmentation for measuring the general energy efficiency of buildings based on their size and function, such as hotels, offices, and the like. "We'd expect that a similar approach would be taken for data centers," he said.

The rating also doesn't take into account other variables, such as climate. A data center located in a cooler part of the world arguably has a built-in advantage over one situated in a hotter climate, in that less mechanically generated cooling is necessary.

However, as with tier level, the EPA decided not to include climate as a variable. Among the reasons, "analysis does not show a statistically significant relationship between climate and energy consumption."

The Energy Star rating system for data centers -- like Energy Star for servers -- is a good first step, though it's certainly a work in progress. The EPA acknowledged this fact. Just as Energy Star ratings are updated periodically for computers and appliances to reflect inconsistencies as well as technological advancements, so too will this rating evolve.

In the meantime, it will set an efficiency bar for data center operators, encourage more organizations to measure power consumption, and (hopefully) improve the efficiency of their operations.

This story, "Energy Star for data centers has bright sides, dark sides," was originally published at InfoWorld.com. Follow the latest developments in green IT and read more of Ted Samson's Sustainable IT blog at InfoWorld.com.

Join the discussion
Be the first to comment on this article. Our Commenting Policies