Uptime chief: Don't believe the PUE hype

Uptime's Kenneth Brill says some companies are misusing the popular metric as a "tool of marketing manipulation"

Astonishingly low PUE (Power Utilization Effectiveness) numbers have become bragging points for some datacenter operators of late, but Uptime Institute Executive Director Kenneth G. Brill cautions the metric is being abused as "a competitive tool for marketing manipulation and misinformation."

In an article on Forbes.com, Brill points out that organizations have been claiming they've achieved PUE numbers of 1.6, 1.2 -- even an impossible 0.9.

[Datacenter operator ADC claims its to have built a datacenter with a PUE of 1.1. Learn more about it here.]

PUE, originally developed by the Uptime Institute and later promoted by The Green Grid, compares how much power your entire datacenter is using -- including servers, CRAC systems, and systems -- to how much of that energy is going toward computing tasks. A lower PUE means a datacenter is operating more efficiently, as the facility is wringing more performance per watt. "A PUE of 2.0 would mean that for every two watts at the meter, only one watt gets to the IT hardware," Brill explains in his article.

Although measuring PUE can be useful for performing "a meaningful health check-up on your datacenter's power and cooling energy efficiency," Brill says, company's claims to extremely low scores should be regarded with some healthy skepticism.

One of the problems with PUE, Brill argues, is that it currently looks only at electricity. " Electricity is only one of a number of data center energy sources. Others include natural gas, steam, chilled water, diesel, free-cooling, and one-pass cooling. So, if electricity alone is counted as an energy source, some sites will appear to be better than they really are," he writes.

Additionally, datacenter operators reporting incredibly low PUE numbers might only be sharing a single score taken at an opportune moment in time. "If averaged for a minimum of 24 hours, a week or, even better, a full year, the number would likely be much higher," Brill writes.

Finally, Brill points out that not all datacenters are created equal; if you want to see how your datacenter's efficiency compares to the competition, "benchmarking is useful only if you are comparing yourself with others with similar power and cooling equipment types, percentages of load utilization, tier design levels, and ZIP codes."


Copyright © 2008 IDG Communications, Inc.