Measuring data center efficiency is not just a way of seeing how well you are doing, but what you could do to increase efficiency. Getting the metrics right by which you measure efficiency is therefore critical to data center planning.
It depends very much on what agency officials see as their particular efficiency challenge – and what they want to get from the data center. Many agencies are focused on reducing their energy costs. Their goal is to use the smallest amount of power necessary to run their IT and cool the servers. Others are more focused on system uptime and application availability and so might be willing to lose a little on the energy use side of the equation.
Power usage efficiency (PUE), a metric developed by the Green Grid, an international industry consortium, has been a major driver for greater energy efficiencies in data centers since it was published in 2007. It represents the ratio of the total power used to run the data center facility, including that for lighting and cooling, to the power used just to run the IT equipment.
The ideal PUE is 1.0. The FDCCI has made a PUE between 1.3 and 1.6 a target for agencies, while also acknowledging that the typical results achieved in data centers so far are usually in the range of 2.0 to 3.0. Another metric, data center infrastructure efficiency (DCIE), is the inverse of PUE.
The past few years have seen a rising interest in factors other than straight efficiency, such as environmental impact. With that in mind, the Green Grid announced in November 2012 that its members had agreed on measurement guidelines and development steps for three new efficiency metrics: Green Energy Coefficient (GEC), Energy Reuse Factor (ERF), and Carbon Usage Effectiveness (CUE).
The new metrics “go a long way in addressing the limitations of the original PUE metric, which only really measures how efficiently the energy consumed is being used,” consortium officials said in a statement. “It doesn’t look at the inputs to and outputs from the PUE, nor at the implications in terms of carbon emissions.”
However, others question the value of PUE and the other metrics, saying they don’t address the larger but related issue of how well data centers deliver on their IT mission. David Cappuccio, who is responsible for data center research at Gartner, said PUE might help an agency reduce energy costs, but it does nothing to help that agency get the most out of its IT resources.
“PUE makes no sense because it has nothing to do with IT efficiency, it just measures the facility,” he said. “I tell people that, if you run the IT shop and want to bring in newer technology because it will improve performance and productivity but also happens to be more energy efficient, then PUE can actually get worse even as the IT gets more efficient.”
Instead, he proposes using a Rack Unit Effectiveness (RUE) metric. Based on the same optimal/actual usage concept behind the PUE, his RUE would measure how much of the available rack space is being used.
This could also be used as a planning tool from a technology refresh angle, he said. If organizations have, say, an 80 percent density in a data center today and want to upgrade half of the racks to current generation servers, they could use RUE to see the impact this would have on the space itself.
“That turns out to be an eye opener for most people,” he said.