The Green Grid, a leader in energy efficiency standards for data centers, has recently updated their PUE (Power Usage Effectiveness) metric that they came out with in 2007 and have outlined the new benchmarks in a white paper titled, “Recommendations for Measuring and Reporting Overall Data Center Efficiency“.

PUE is defined by the total data center energy consumption or power divided by the IT energy consumption or power. So, the lower the resulting variable, the higher the energy efficiency of the data center because more of the total energy is being used for its intended function – to power the IT equipment – rather than to cool the servers, provide lighting, support infrastructure, etc.

The PUE standards have been updated recently to account for the  loopholes that several companies had exploited to lower their PUE value (these abuses were reported on Forbes.com and InfoWorld.com, among other places). As a result of these considerations, new standards such as including a weighted measure of different types of source energy (not just electricity but natural gas, fuel oil, other fuels, district chilled water, district hot water, and district steam) and a requirement to take readings from energy meters over a 12-month period (rather than over one 24-hour duration and extrapolating for the entire year).

The new recommendations are definitely worth a read for those involved with data centers. There are a few important points to keep in mind about Version 1 of the new PUE standards:

  1. Dedicated data centers – the new standards are limited to data centers which are located in a building where the power, cooling, lighting, and support infrastructure are devoted exclusively to them. An update to the current document will include guidelines for measuring PUE in mixed-use facilities.
  2. Report PUE with subscript – there are four categories of PUE (0-3), depending on how energy consumption or power is measured. PUE3 is rated the highest standard because the energy consumption reading is taken at the point of connection of the IT devices to the electrical system, not at less accurate locations. PUE0, the most generic category, can not be used when source energy includes types other than electricity. PUE must always be reported with the subscript of the category used, and the eventual goal is for PUE3 to become the industry standard of data center efficiency measurement.
  3. Must subtract non-IT equipment power use – Categories 0 and 1 measure IT energy at the UPS output location. According to the white paper, “If there is non-IT equipment supported by the UPS system (e.g., CRAC, CRAH, In-row coolers, etc.) it must be metered out and subtracted from the UPS Output reading”.
  4. Energy sources are weightedSee the white paper for a table of the different factors by which source energy types are weighted.
  5. Heating nearby homes doesn’t count – A trend in data centers has been to use excess energy generated from cooling the server rooms to heat other buildings (for example, a data center in Zurich heats nearby homes). These efforts to improve overall energy efficiency do not lower the PUE score in any of the categories because PUE is defined as an efficiency measurement of only the dedicated data center itself. According to the report, though, “currently there are on-going industry efforts to define a metric that could be used to account for this beneficial use, but it is specifically excluded from PUE.”

According to an article on GreenBiz.com, billions of dollars are wasted a year worldwide to run and maintain servers that sit on racks unused. Hopefully increased standards for energy management in data centers, such as the PUE metric, will help us be more mindful of how energy is used (or misused) in our IT hardware hubs.

Post to Twitter Post to Delicious Post to Facebook Post to Reddit Post to StumbleUpon