The virtual summit takes place Wednesday, Sept. 27 from 1-3 p.m. ET. fnPrime members can register for free
Bring your questions and get answers from Joan Stein, nationally recognized ADA expert, in this interactive virtual session
Modularity allows building owners to increase capacity incrementally, spreading the cost of capital over several years. But that makes sense only if there's enough space to allow growth to spread out over time, says Terry Rennaker, vice president of Skanska.
The challenge becomes building an infrastructure that is flexible enough for the 18-month life cycles of information technology. For instance, a state-of-tomorrow data center built in 1996 for Microsoft could handle as much as 1.5 kilowatts per rack. At the time, some thought that density capability was wasteful overbuild. Today, eBay's flagship data center in South Jordan, Utah, supports power densities up to 30 kilowatts per rack.
Rennaker admits designing data center infrastructure is far more challenging than building an office building. "People's needs do not change dramatically from generation to generation," he says. "We can be pretty sure that an office building will still be useful 20 years from now."
The problem becomes how much infrastructure to build into the space so that power usage effectiveness (PUE) is optimized. Obviously, if a data center has 10 megawatts of capacity but only has 1 megawatt actually plugged in, electricity is being wasted.
"Electrical equipment runs better when it's fully loaded," says Rennaker. "So if we add power generation incrementally, we can keep operating at the higher efficiency levels."
PUE and energy efficiency certainly are fueling the shift to modular thinking. "In modular data centers, the infrastructure (power and cooling support equipment) will be scalable, so that they can more closely match the IT equipment load, and thus operate more efficiently," says Kevin McCarthy, associate vice president of AECOM.
To this end, modular UPS systems and such elements as in-row direct-coupled cooling become the building blocks of modular data infrastructure, according to McCarthy.
The life-cycle cost of owning a modular data center should be examined, says Rennaker. "What is the total cost of ownership, including financing, capital outlay, operating expenses and utility expenses, to own this data center?"
A detailed cost analysis will show cases where a modular approach isn't the best strategy. "In general, the total cost of the full build out will be more expensive for a modular data center," says Sewell. "Designing in modularity may not make sense if the apps are all of the same criticality and growth is not an issue."
Facility managers also will want to look at the ripple effect that infrastructure construction cost has on rack requirements. Schlattman admits there currently is not a benchmark for per rack costs, but he feels that is a good way to evaluate modular alternatives. "After you do the math, you may find one modular solution costs $60,000 per rack, while another is $50,000 per rack," Schlattman says. "If as an end-user you need to support 50 racks, that's a significant cost difference."
Another issue to evaluate is what the data center industry calls "white space." In this area are all the supplementary elements and premium sizes that don't fit within the standardized rack width, depth and height.
Other areas to consider:
While modularity may not be an option for every data center, it can save first costs, increase operational efficiency and provide future flexibility, all pluses in mission critical facilities.
Rita Tatum, a contributing editor for Building Operating Management, has more than 30 years of experience covering facility design and technology.