This month the Environmental Protection Agency released its Energy Star Rating for data centers and there has already been much conversation on the subject. An intelligent summary and critique can be found in Bob McFarlane’s article, “Energy Star for Data Centers is First Definitive Green Data Center Rating,” on the Searchdatacenter.com Data Center News.
The gist of Mr. McFarlane’s article is that despite some flaws with the PUE metric, this is still a major step forward in providing both a benchmark basis and incentive for efficient design and operation of data centers. While the data collection requirements will not prevent potential abuses from inefficient IT operations actually driving down the PUE – you can make yourself a smaller PUE (quotient) by either lowering the dividend (total energy including mechanical infrastructure and electrical losses) or raising the divisor (IT load, as in poor utilization) – I think we can assume at the very least most Energy Star applicants will want to avoid the irony of operating with servers that don’t qualify for good Energy Star ratings. This article did point out that the Energy Star metrics were excluding geographic variances, which would theoretically place cities in warmer climates at a disadvantage in terms of capitalizing on economizer (free cooling) benefits.
Logically, that observation makes perfect sense – a data center in Billings, MT is going to have access to many more free cooling hours a year than a data center in my home town of Austin, TX. However, of the 100 plus data centers than contributed data for the benchmarking study, there was a proportion that utilized some form of economization and there was no statistically meaningful geographic bias or variation in the resultant PUE’s. More than anything else, that data suggests that economizers apparently are not being utilized very effectively; so for the immediate future, there is more room for overall improvement than there is for certain operators wresting an unfair geographic advantage.
In the final analysis, empirical data will be more defensible than speculative logic as data center operators and designers begin jockeying for position. As a matter of fact, achieving Energy Star Rating is by definition a moving target, since it requires a year’s accumulation of energy use data that places a date center in the 75th percentile or higher of applicants, and we should rightfully anticipate that bar will continue to be raised from year to year. Since that is already a temporally-based moving target, I suspect that it would not be unreasonable to see the EPA make a geographic adjustment somewhere down the road, if subsequent data demonstrated that was necessary.
On the flip side of that, however, site selection is an integral factor of the overall data center design process, so maybe a good future compromise might be to grandfather in some geographic compensatory scales, but keep a single scale for new construction. Regardless, for companies like Chatsworth Products, Inc., who offer solutions to facilitate designing and operating highly efficient data centers, we applaud anything that raises the general industry interest in energy efficiency. Ian Seaton, Technical Applications Development Manager