Is the Power Usage Effectiveness (PUE) metric the best way to measure data center efficiency?
I'm not the first one to ask that question, and I probably won't be the last. However, it does seem that IT executives are growing overly comfortable with PUE at a time when a growing chorus of voices are warning against its enshrinement as the standard for energy use.
At the moment, the EPA is firmly in PUE's corner. The agency is on the verge of releasing a Portfolio Manager tool that will be used to gauge individual data centers' qualifications under the Energy Star program. It's the same system used to track use at banks, hospitals, offices and other commercial buildings, and it uses PUE as its guiding principal.
The problem, according to critics, is that PUE calculates efficiency as a ratio between total consumption and consumption of strictly IT-related equipment -- a lower number indicating greater efficiency. It does not take into consideration overall data loads, system performance factors or other measurements to indicate energy consumed as related to the amount of data being processed or the speed at which it does so. It also does not factor in cooling requirements for IT equipment, allowing data centers in colder climates to achieve lower PUE ratings than those in warmer areas.
Other countries are already looking beyond the PUE to provide a more complete picture of a data center's overall environmental impact. Australia, for example, is preparing to tweak its National Australian Built Environment Rating System (NABERS) program to the IT industry by looking not just at power consumption, but things like water usage, waste practices and even building materials. The agency that monitors the NABERS system, the New South Wales Department of Environment, Climate Change and Water (DECCW) is canvassing data center executives to determine how the metric should be configured.
Even the Green Grid, which developed of the PUE system, is quick to admit that it does not provide the entire picture of data center efficiency. The organization is at work on a new set of metrics that take into consideration factors such as local power generation and the availability of "free" cooling from local water sources or cooler outside air.
But adding too many variables to an energy-efficiency rating can have drawbacks as well. As data centers are allowed to claim an ever-widening field of extenuating circumstances to couch their particular scores, direct comparisons become difficult, if not impossible, and the ability to draw any conclusions about overall energy efficiency is compromised. As it stands, less than 20 percent of data centers feel they achieve a PUE ratio of 2.0 or less (the average is 3.0), according to the Digital Realty Trust. But if those measurements are skewed due to outside ambient air temperature or other factors, expect to see better PUE reporting, even if energy consumption remains the same.
The whole business points to the fact that Mother Nature is an extremely cagey individual. Changes in the environment, whether man made or natural, tend to have both positive and negative consequences, making it difficult to choose the correct course of action.
One thing is clear: Energy costs are high and likely will remain so for the foreseeable future. Any action you take to bring those costs down will most definitely improve your bottom line, no matter what number you write on a government form.