IO Conducts PUE Faceoff: Modular vs. Raised Floor

IO-Anywhere-Modular-System

A row of IO Anywhere data center modules at the IO Phoenix facility. (Photo: IO)

Which delivers the best energy efficiency: raised-floor space or a modular data center? It's a hard question to answer in a satisfying fashion without comparing apples and oranges. What about a comparison of modules and raised-floor space within a single facility, using the same building envelope and chiller plant, with the data reviewed by a third party?

IO had exactly this situation in its IO Phoenix facility, which is split between an initial phase of raised-floor data center space and a second phase filled with the company's IO.Anywhere modules. So IO asked the local utility, Arizona Public Service (APS) to review 12 months of data for both environments and calculate their respective operating costs and Power Usage Efficiency (PUE).

The result? APS found the modular design offered significant improvements in efficiency and economics. The utility said the raised-floor area within IO Phoenix had a PUE of 1.73 for 2012, while the modular data center environment had a PUE of 1.41. That difference translates into an annual savings of $200,000 per megawatt of average IT power for customers using the IO.Anywhere modular build-out instead of the raised floor space at IO.

“Our calculations did show that the IO.Anywhere modular data center uses less energy than a traditional data center build-out, at least in the case of this IO data center” said Wayne Dobberpuhl, APS Energy Efficiency Program Manager. “Moving forward, we are working with IO to establish the right baseline for assessing the appropriate rebate for this efficiency work under our Solutions for Business program.”

Customer Analysis Focuses on Operating Costs

IO has a clear perspective in this debate, as the company readily admits. "Our actions have long spoken to our faith in the modular design," said Patrick Flynn, the Lead Sustainability Strategist for IO.

But Flynn said the data will be "exceptionally valuable" to enterprises pondering the best deployment option and evaluating modular units. He noted the importance of having Arizona Public Service provide an independent analysis of the potential energy savings.

Operating costs have become a focal point in comparisons between traditional hot aisle/cold aisle data centers and modular offerings like IO.Anywhere, that are built in a factory using repeatable designs and can be shipped to a customer premise or provider facility. A 2012 study from 451 Research found that modular data centers are usually cheaper to build than traditional raised-floor space, but said the economics of the two deployment models were complex and didn't stake out a position on the OpEx economics of the two models.

In analyses with many moving parts, proponents of each approach have found support in PUE, the energy efficiency metric popularized by The Green Grid that compares a facility’s total power usage to the amount of power used by the IT equipment, revealing how much is lost in distribution and conversion. The average PUE is about 1.8.

Advocacy For Both Designs

Factory-built designs from Microsoft, SGI and others have documented the ability to deliver PUEs of 1.2 or lower using containers and modular form factors in a free cooling environment. But those customized hyperscale facilities are alien turf to many enterprise customers, who yearn for an apples-to-apples comparison of how modular designs perform in data centers that look like theirs.

Meanwhile, wholesale data center providers argue that their scale allows them to deliver data suites in multi-tenant facilities at attractive PUEs. That's why Digital Realty began publishing PUE audits to document energy gains in its facilities back in 2010. Both Digital Realty and wholesale rival DuPont Fabros Technology report PUEs around 1.3 in their most efficient facilities.

IO saw an opportunity to use its Phoenix facility to demonstrate the relative cost and efficiency performance of the competing models in the same footprint. Flynn says the modules gain their efficiency advantage partly from their compartmentalized design, which manages airflow more efficiently than the high-ceiling colo area at IO Phoenix. Flynn said a key piece of the difference is the IO.OS data center management software, which is implemented more extensively in the modular environment than on the raised floor, and allows more granular management of cooling and energy usage.

"The portion of PUE above 1.0 denotes energy not going to IT equipment, and that’s where efficiencies can be found," said Flynn. "We’ve reduced this portion from 0.73 down to 0.41 in our switch to the IO.Anywhere modular data center, which we recognize as a 44 percent reduction in energy spent on infrastructure versus primary business applications."

Making PUE More Useful

IO used PUE as its benchmark for its comparison, but recognizes that the metric itself can prompt debate.

“PUE is often criticized," said Flynn. "It has some limitations that people enjoy griping about. We recognize that PUE has an important place in customer assessments of a data center’s cost effectiveness and environmental sustainability. Part of our job at IO, therefore, is to validate PUE in actual deployments today, and to continually improve data center performance. I think it's extremely valuable for what it does, but it's  certainly very difficult to compare one data center to another based on PUE alone."

Flynn and IO are among those pushing the value of real-time metrics on energy use and efficiency. Intel has also been making the case for real-time data as a differentiator, offering tools to allow companies or DCIM software vendors to incorporate into their tools. Meanwhile, Facebook has created a live public dashboard updating the PUE for its facilities in real time.

"Real-time PUEs are one way we can push efficiencies further," said Flynn. "I think it's the next step in making PUEs more useful. We can break it down so end users can benefit from it (in a multi-tenant environment)."

That's why real-time data is a key focus for IO.OS software, which Flynn says can offer current readings on PUE and efficiency for customers with a variety of configurations. It has begun implementing this real-time PUE in its IO New Jersey data center, offering multiple views into IT efficiency.

To accompany the Arizona Public Service findings, IO created a graphic view depicting the results in detail, along with its views on the advantages of modular data centers:

io-PUE-Infographic-470