With concerns over energy costs a perennial issue, and environmental sustainability only likely to become more important in the future, minimising power wastage is a priority not just for data centre operators but their customers also.
Power Usage Effectiveness (PUE) is an important and recognised metric for understanding how efficiently a data centre is delivering energy to the Information Technology (IT) equipment on which its services are based. Stated simply as the ratio of total energy consumed by a facility to the energy expended by its IT equipment, a PUE rating is easy to understand and, at first sight, easy to calculate. It gives a numerical rating for how much power, relative to that needed for its IT equipment, a data centre is expending on other functions, in particular cooling.
Many data-centre operators now quote their PUE ratings in marketing messages, so keen are they to exploit their efficient use of power as a measure of their worth. The closer to the figure of 1.0, the better is the data centre minimising power wastage. But with that new visibility comes concern that PUE ratings may not be a fair way to compare the energy efficiency, let alone the overall productivity or effectiveness, of one data centre to another. Is there a danger that the aggressive selling of impressive PUE ratings by some operators may reduce the credibility and therefore the usefulness of the metric as a whole?
As always, the buyer must beware. Sometimes a quoted PUE rating raises more questions than its numerical simplicity appears to answer, especially if it purports to reach, or even exceed, the ideal but unrealistic limit of 1.0.
The PUE metric, first published in 2007, was developed by The Green Grid Association, a global non-profit, open industry consortium of end users, policy makers, technology providers, and utility companies working to improve the resource efficiency of information technology and data centres throughout the world. Since then, The Green Grid has continued to refine the metric measurement methodology with collaborative industry feedback
In fact the PUE metric was not originally intended as a means of comparing energy efficiency between different data centres. Rather it was devised as a tool to enable individual data centres to monitor their own efficiency over time, taking into account the inevitable changes in both equipment and operating practices that would occur during its operating life.
Given the eagerness of data centre operators to publicise PUE ratings they calculate to be lower than their competitors, the Green Grid has over time produced a set of rules and guidelines regarding the process that organisations should follow when making public claims. These go some way to helping decision makers judge how a particular PUE rating may have been calculated and what weight should be attached to it.
The Green Grid classifies publicly reported PUE results using four categories: Unrecognised, Reported, Registered, and Certified. The first two categories are essentially self-assessments by the reporting organisation which may claim that it followed the Green Grid's measurement recommendations and nomenclature guidelines (Reported) or make no such claim at all (Unrecognised). The Green Grid makes no comment on results presented in either of these categories.
Organisations with Registered and Certified results have provided the Green Grid with additional data about their measurements so that the Green Grid can analyse overall industry performance and energy-efficiency trends over time. Only Certified results, which have been achieved using the most stringent data-reporting requirements, qualify for inclusion in any program created by the Green Grid to award or recognise data-centre efficiency.
The first element to consider therefore when considering a claimed PUE rating is what classification can be applied to the results. In the Unrecognised classification, a data centre operator may choose to ignore some or all of the measurement and calculation recommendations set out by The Green Grid.
For example, the Green Grid awards no credits or percentage points for generating energy or recovering waste heat. An operator that generates some or all of its electricity requirements using renewable resources such as solar or wind energy may be making a valuable contribution to environmental sustainability, as well as creating opportunities for long-term cost reduction for itself and its customers, but these are irrelevant to the calculation of PUE as recommended by the Green Grid.
Another element to consider when comparing PUE ratings is how the various electrical loads in a data centre operation have been assigned in their calculation. The apparently simple separation between IT and non-IT equipment, essential for an accurate calculation, is in practice rather complex.
The Green Grid produces a comprehensive list of which equipment should be categorised as IT-only, and therefore appearing below the line of the equation used to determine PUE, and which equipment counts as ancillary, to appear above the line. In most cases, the classification is straightforward: computing equipment including servers, processing elements, storage arrays, networking equipment, and KVM (keyboard/video/mouse) switches go below the line as do, monitors and workstations or notebook computers used in the network operations centre (NOC).
Along with these on the top line of the equation goes all other energy-consuming equipment needed in the data centre. This includes: cooling equipment such as chillers, cooling towers, pumps and air-conditioning units; lighting, and power distribution equipment including UPS (uninterruptible power supply) systems, switchgear, generators, power distribution units, and batteries.
However as technology evolves, the demarcation lines between the various types of equipment become blurred. Some cooling elements such as pumps, refrigeration, blowers and heat exchangers are now to be found within the IT equipment itself. To which part of the equation should the power associated with them be assigned? Maintaining consistency of calculation is an ongoing challenge for the Green Grid.
As well as continuously updating its classifications of what does and does not constitute IT equipment, the Green Grid also recommends the use of energy-measuring techniques throughout the data centre so that accurate calculations can be made. Measuring the total power expended by a facility at the utility meter is simple and inexpensive but does not make for accurate assignment of power losses for the purpose of PUE calculation. Installing measurement apparatus throughout the data centre can be expensive, and introduces a further layer of complexity into the calculation of results which must also be taken into consideration.
When taking into account all of the inevitable energy losses in a data centre between cooling, lighting and measurement as well as running the IT equipment itself, it is impossible, following the recommendations of the Green Grid, for a Certified data centre to attain a PUE rating of 1.0.
Conclusion
PUE is a useful and widely accepted tool for measuring the energy efficiency of a data centre. However, it is best applied to monitoring the energy efficiency over time of an individual data centre as its contents and purpose evolve to accommodate changes in technology and customer demands. If used as a means of comparing one data centre to another, which may have a different configuration, a different purpose and located in a different environment, care must be taken to analyse how its rating was calculated and whether one is really comparing apples with apples.
Schneider Electric White Paper #158, Guidance for Calculation of Efficiency (PUE) in Data Centers, is available for free download at www.apc.com/whitepapers
