EPA ENERGY STAR to Offer New Rating System for ICT Facilities

The 7 June release by the US Environmental Protection Agency (EPA) of its building Portfolio Manager will include new features for data centers, both as standalone buildings and as 'spaces' within larger buildings. One of the features is a new ENERGY STAR® rating systems for data center buildings, but not for data center spaces. (A data center space will contribute to the overall ENERGY STAR rating of its building.) Here is how the EPA explains its rating methodology:

EPA used statistical regression analysis to identify the operating characteristics that explain the variation in PUE among data centers. Of the numerous variables analyzed, only the annual IT energy consumption was found to be statistically significant in explaining the variation in energy use; data showed that facilities with higher IT energy consumption have lower PUE values, on average. The final regression model predicts a PUE value based on the IT energy. For a given facility the actual measured PUE can be compared with this prediction. A facility whose actual PUE is lower than the predicted PUE is doing better than average. This performance level (actual compared with predicted) is mapped to a 1-100 scale such that one point represents one percent of the population.

The basic concept, as I understand it, is that a lower-is-better metric, PUE, will be mapped into a higher-is-better metric, the ENERGY STAR building rating, correcting for a bigger-is-better bias. It is important to understand that there is not direct mapping of PUE into the rating system. Instead, the EPA rating table maps an "energy efficiency ratio" (actual PUE/ predicted PUE) into a rating.

According to the EPA, "Buildings achieving a rating of 75 or higher and professionally verified to meet current indoor environment standards are eligible to apply for the ENERGY STAR [label]." It is difficult to predict what PUE might earn a 75+ since the ratings are based on the EPA's "energy efficiency ratio", not a facility's actual PUE.

The "energy efficiency ratio" approach suggests a data center with lower PUE will not necessarily get a better rating than another with a higher (worse) PUE . A smaller and less efficient data center can get a good rating if it is still relatively efficient for its size. Conversely, a very large and more efficient data center could get a lower rating if it is not keeping up with its mega-brethren.

The EPA also states "The PUE generally ranges from 1.25 to 3.0 for most Data Centers". and the rating system is based on a mean PUE of 1.9. This is lower (better) than other reports. The difference may be the result of EPA working with industry groups already committed to Green ICT like the The Green Grid. I suspect the PUE range for the 61 data centers analyzed is better than for the industry as a whole and some data centers might find they get poorer (relative) ratings than they might have anticipated when they apply the ENERGY STAR Building Portfolio Manager to their facilities.

The most interesting outcome of the EPA's analysis of existing data center data was finding that climate had no impact on the cooling load:

Climate is one characteristic that was examined closely. EPA found no statistically significant relationship between heating and cooling degree days and PUE, which was initially considered to be surprising. However, upon further review, it was determined that the energy required for cooling a data center is dominated by the high internal loads generated by the IT equipment, and that climate has a relatively low contribution to the building cooling load.

Internal loads are much higher than Climate loads...Reportedly a 10 to 1 ratio, much higher than commercial buildings

I'll join the EPA in calling result "unexpected". My initial reaction was to assume that few of the reference set data centers use air-side cooling, for which climate can play a major factor. The finding also appears to contradict the anecdotal experience of trying to use PV solar to power a data center in a sunny but hot climate.

The EPA tells me, "In fact, there were 27 out of 61 stand alone data centers that reported economizers (44%)." What is going on here? The EPA hypothesizes, "Economizers do not appear to be working properly. This result has been observed in other building types, where economizers can be disabled or prone to failure." This reinforces the critical point that Green ICT is as much about operations as it is about technology.

Update 2010.06.11
Vertatique uses a much richer definition of ICT facilities than just conventional data centers. The EPA just confirmed to me that ENERGY STAR for Data Centers can apply to other ICT facilities that have similar characteristics. These include independent UPS and cooling for the ICT gear. The specific example I used was a media industry facility containing a mix of IT, telecommunications, and video equipment rather than just IT servers and related gear.

Update 2010.08.07
See the results of our ENERGY STAR for Data Centers Portfolio Manager tests.


Quoted material is from EPA documents ENERGY STAR® Data Center Infrastructure Rating Development Update (29 Sep 2009), New Data Center Energy Performance Rating (19 May 2010), Technical Methodology for Data Center ENERGY STAR® Performance Ratings (Jun 2010), and email exchanges.