How to Benchmark Data Center Energy Costs

In the spring of 2007, UPS’s Ben Swanson and Joe Parrino attended a conference on the growing problem of data center power consumption. One suggested remedy was to benchmark and analyze the power flowing through the data center. So after the conference, Swanson, the facilities department manager, and Parrino, a data center facilities manager, profiled their two main data centers to learn what they could do to increase their energy efficiency and, ultimately, save money.

More on CIO.com

Five Ways to Find Data Center Energy Savings

Good Incentives Boost Data-Center Energy Efficiency

Signposts on the Road to Data Center Energy Savings

They found one crucial area to improve: their 65 computer-room air handlers. Parrino and his team found that some of their power distribution units had perforated tops. “We were losing all sorts of air through the tops of the cabinets,” he says, which meant that some of the cool air circulated by the air handlers was being wasted by cooling the PDUs. They restricted the air flow through the cabinets, tested to make sure they wouldn’t overheat, then turned 24 air handlers off.

The move has saved UPS 1.6 million kilowatt-hours and about $124,160 per year. And that’s not all—they also benchmarked their mechanical cooling system and were able to reduce its energy use, saving an additional $100,000 annually. The mechanical plant of these UPS data centers consumes about half what a typical data center mechanical plant consumes, according to their benchmarking study, Parrino says.

Despite the potential for savings, however, most IT departments have not rushed to benchmark data center energy efficiency, primarily because there’s no incentive to measure it, says Forrester analyst Doug Washburn. He notes that only 11 percent of IT organizations are responsible for paying their energy-related operating costs (the power bill typically goes to the facilities group). But that’s changing: More CIOs are being asked to reduce energy consumption, Washburn has found, and, as a result, new metrics—though they come with caveats—appear to help pinpoint which areas of your data center can be optimized.

Two Ways to Benchmark

One such method of measuring energy efficiency, developed recently by the Uptime Institute, a data center research organization, is Corporate Average Datacenter Efficiency (CADE). This calculation multiplies the efficiency of one’s technology by the efficiency of the physical facility. Uptime calculates IT efficiency by multiplying a data center’s IT asset utilization rate by the energy efficiency of the servers. It determines the facility efficiency by calculating the amount of space used and multiplying that by the energy efficiency of the building. The higher the CADE number, the more efficient your data center. This method tends to be best-suited for larger businesses since those usually have the automated monitoring tools to measure server utilization, according to Uptime Institute’s Executive Director Kenneth Brill.

Another method—Power Usage Effectiveness (PUE)—was developed by The Green Grid, a group of technology companies collaborating to improve energy efficiency in data centers.

The PUE is calculated by dividing the total utility load (the power coming into your facility) by the total IT equipment load (the power consumed by switches, routers, servers and related gear). In this instance, the lower the PUE, the better.

Forrester’s Washburn warns that while these methods can be effective, the measurements required to calculate the benchmarks can be challenging to obtain. “Because most organizations do not pay for their energy-related IT expenditures, many are not measuring energy consumption—especially on a granular level—which is required to actually calculate your data center energy efficiency metrics,” he says.

Your ROI May Vary

Although the emergence of industry-accepted benchmarks means you can be more certain you’re measuring the right things, Washburn also notes it’s usually larger businesses that benefit from benchmarking. The bigger the data center, the more energy it consumes and the more likely it will be that you’ll have an incentive to reduce consumption. “For smaller businesses, the financial impact may not be as significant, and you can direct your time elsewhere.”

What’s more, there isn’t enough publicly available data yet for companies to benchmark against each other, so it’s hard to know how efficient you can really get. If you try to compare your energy consumption with another company, peers in your industry might be a good start, says Washburn. But a more relevant comparison would be an organization with a similar-sized data center and similar server, storage and network needs.

Christian Belady, principle power and cooling architect for Microsoft Global Foundation Services and Microsoft’s representative to The Green Grid, agrees that there are limits to these metrics, specifically for smaller businesses. But he thinks even small organizations can still benefit by benchmarking on a smaller scale. Belady says that once CIOs decide to benchmark their energy usage, they should consider benchmarks such as The Green Grid’s PUE or any other metrics that may make sense for driving the right behavior for their business. “The industry is moving so fast that you can’t wait for everyone to have the perfect definition of these metrics,” he says.

One key to UPS’s benchmarking success, Parrino and Swanson believe, is the support and trust from their CIO, David Barnes, to make these decisions. “You need to constantly strive to challenge yourself and be constructively dissatisfied with your own operation,” Swanson says.

Whether these metrics come from vendors or industry organizations, Washburn believes that they’re a good starting point for benchmarking. “At the end of the day, in most cases you’ll be able to use the metrics regardless of which method you chose.”

Whether these metrics come from vendors or industry organizations, Washburn believes that they’re a good starting point for benchmarking. “At the end of the day, in most cases you’ll be able to use the metrics regardless of which method you chose.”