Developed by The Green Grid, power usage effectiveness (PUE) is a very useful metric for measuring the energy efficiency of a data center. PUE is calculated by dividing the overall power entering the facility by the power consumed by the IT load within it — the target being to get as close as possible to a PUE of 1.0, as this would mean the data center is 100 percent efficient.
Designing data centers for low PUE and measuring actual performance is becoming increasingly important, as operators seek to minimize their environmental impact and reduce their operating costs. One of the results of this trend is that it is now nigh on impossible to find a data center vendor that does not brag about its products’ exceptional PUE performance. But rather than accepting these marketing claims at face value, here are five important questions you need to ask to make sure that you’re getting the full energy efficiency story:
- Where is the power being measured?
The result of a PUE calculation can be highly variable depending on where the incoming power and IT load are being measured. For a data center, the overall facility power should be measured at the incoming main low voltage distribution board (MLVDB) and the IT load at the rack PDUs. However, some vendors measure their IT load at the sub distribution boards, thus achieving a better looking PUE because they have ignored all the cable loss from the low-voltage distribution boards to the racks. The table below summarizes the four recommended categories for the measurement of PUE. Of course the ultimate PUE calculated and claimed for any given facility will vary depending on which measurement category is used.
- Are all losses taken into consideration?
When calculating a data center’s PUE, you need to make sure all losses are taken in to consideration. Unfortunately, some vendors omit some losses from their calculations in order to achieve a much better efficiency figure, so you need to be sure they are including all losses in the switchgear, cables, lighting and UPS, as well the increased load on the cooling systems dealing with heat generated by the sun outside or even by the people inside the facility. Only then will you arrive at a realistic PUE for your facility. For more on this, here is a very interesting article on Google’s approach to measuring PUE in their data centers.
- Is PUE a one-time calculation?
The amount of energy a data center needs will vary over time depending on a number of factors. For example, an increased reliance on cooling systems during warmer weather will mean that PUE values are generally higher during the summer than the winter. So rather than making it a one-time calculation, PUE ideally needs to be calculated over the entire year to find the annual average as well as the annual peak. And from an operational cost planning perspective, accurately calculating this worst case figure is actually more important than knowing the best case scenario.
- Do local weather conditions affect PUE?
Most vendors like to claim PUE values of 1.1 or 1.2 for their data centers, and this is of course entirely possible if the facility is deployed in a cold environment like the Nordics where free air cooling can be used very effectively. However, if you are building a data center in the tropics, your best case PUE is going to be around 1.5 to 1.6, so you need to make sure that the PUE value that any vendor is offering is based on the location where it is going to be built. Ask if they have done any modelling based on the weather patterns of the actual site location. Even if they have not built one there before, peak and average PUE can be accurately determined using statistical climate data (temperature and humidity) with simulation software.
- Will a vendor guarantee the PUE they claim?
With all of the above said, and given that it is relatively easy to manipulate PUE calculations so that the resulting values look better than they really are, possibly the most important question you can ask any vendor with regards their PUE claims is whether they will guarantee that level of performance over time?
Armed with these five questions, you should be able to cut through the marketing hype and compare competing vendors on a relatively level PUE playing field. Of course, the economics of PUE is another subject altogether — the impact on PUE, for example, if powering a facility from the grid or with diesel or both, or the impact on PUE of increasing a facility’s redundancy level. That will have to be the subject for another article.
Srikanth Murugan is global director sales engineering at Flexenclosure, a designer and manufacturer of prefabricated data center buildings and intelligent power management systems for the ICT industry. He has spent nearly two decades working in the telecommunications industry around the world. For the last three years, he has led the sales engineering function at Flexenclosure, responsible for designing bespoke customer data center solutions.