# Data Center Efficiency Calculator – A Tool for Modeling Your Current and/or Future PUE

This audio was created using Microsoft Azure Speech Services

With all the buzz about data center efficiency, and measuring PUE, this tool provides a model to help customers estimate what they can expect for their annualized PUE, and helps them quantify how much PUE can improve when you take specific actions or select different architecture choices.  Parameters include  data center size, expected IT load, power & cooling architecture, and other selectable  data center characteristics.  The results show the overall PUE, and several charts, including PUE & DCiE (inverse of PUE) curve at varying loads, a breakdown of the power & cooling losses, and breakdowns of the energy costs.  So how does it work?

This calculator is based on a patent pending modeling technique for aggregating device-level data and inputs to estimate efficiency.  It looks at individual physical infrastructure components (power, cooling, lighting, auxiliary) and the associated losses (or wasted electricity) from those devices.  For each device, we identify 3 types of losses:  fixed, proportional, and square.

The fixed loss is sometimes referred to as “no load loss”, and I think a lot of people relate to this. It’s what you have to pay for immediately upon plugging the UPS, or the cooler, or the chiller or the pump into the electrical supply.  Once you plug it in, something is going to start drawing power and that “something” is called fixed losses; and those losses don’t change with load, it doesn’t change with anything; you just inherit it the minute you energize it.  Examples include: auxiliary power supplies, communication interfaces (including front panel LCD and LEDs), cooling fans.

The proportional losses increase as a proportion of the load that goes onto a UPS, or chiller, or CRAC.  Most people can relate to that… the more load on the system, the more power it draws.  Examples include the switching losses from transistors, and conduction losses of semiconductors and rectifiers.

Square losses are a little harder to understand.  They become significant at higher loads, for example a transformer; Square losses are the result of a physics law; and what it says is that you lose power by the square of the current that runs through a particular conductor; Conductors such as power wires, bus bars, and circuit breakers produce losses that are proportional to the square of the load.

When you take the 3 losses together, you end up with the total loss of that device.

So where do we get the loss data?  Well, a lot of the data in our tool is measured data, some is estimated.  If we had a product we could measure, we measure it.  In fact, if you were to look at our website, you’d see measured efficiency curves for many of our UPS and cooling systems. Some data like for pumps and chillers come from 3rd party published data sources, etc.

So how do we take measured data and turn it into loss parameters?  Well, if you measure a UPS immediately upon plugging it into the wall, you know that it’s going to draw a certain amount, that’s the fixed losses; then you start loading it up and measure the power draw at the varying loads.  From these data points, we can form a pretty precise curve; and then from that curve, we can deduce what the 3 parameters are by trending the line with a 2nd order equation.

Another point I want to make regarding this model is that the it accounts for circular energy flows. Think about this example: If you want to put the fans of air handlers on UPS power, that increases the load on the UPS; but more load on UPS means more heat that the fan needs to remove; if that fan is a variable speed fan, it’ll spin up faster to get rid of that heat, which then becomes more load on the UPS, and what you have here is a never ending loop; Our model deals with these loops so that you are able to do “what-if’s” with scenarios like that.

Here are a couple of our white papers, Electrical Efficiency Modeling for Data Centers (white paper #113), & Electrical Efficiency Measurement for Data (white paper #154), which go into the methodology in more detail.  These are available on our website.

Keep in mind, this is based on a “typical” data center, and differences in sizing and efficiency of specific devices used will cause the actual result to vary.  Assumptions for our sizing and losses are right in the tool, just click on “assumptions” in the upper right.

We’ve had many customers use this tool, and have gotten lots of great feedback on how close the modeled results come to actual measured PUEs.  I hope this tool can be helpful for your next data center project.  I’d love any feedback you have on the calculator – any experiences you can share in using the tool, any questions on data, etc.  If you have general comments or questions about data center planning tools, I welcome those too.

Happy PUE modeling!  https://www.apc.com/tool/?tt=6