Building Operating Management

CFD’s 3-D Detail Can Help Avoid Stranded Capacity





In building a 3-D diagram, CFD modeling is detailed enough to help data centers avoid "stranded capacity" that wastes resources and benefit from its predictive ability.

CFD modeling requires the taking of measurements, which includes determining how big the room is, how high the floor is raised, how tall the racks are, and so forth, explains Koomey.

To do CFD modeling, sensors are put around the data center floor to track air flow and temperature, which allows the CFD tool to build a comprehensive model. It's "a three-dimensional color diagram of the floor so that you can see air flow and where heat is on the floor," Cappuccio says. "You can then make decisions about where to reposition equipment and redirect air flow." New software tools monitor power consumption, heat generation, and air flow, and temperature in near real time at the device and rack level. For a typical data center, this means tracking thousands of data points to understand exactly what is happening on the data center floor.

Gathering data points for air flow and temperature at the racks can take a few weeks, depending on the facility, but once the initial model is created it is simply a question of adding data. Every time a change is made to the data center, a change has to be made to the model. "The model exactly replicates the data center at any point in time," Koomey says, and the model allows for determining the impact of hypothetical scenarios before carrying them out.

Stranded Capacity

Given the way that data centers evolve, cooling and power infrastructure in many is fragmented. The result is stranded capacity — space that cannot be used — in data centers. Wade estimates that up to 30 percent of capacity is stranded in most data centers, and is generating no financial return for the company. About 20 percent of the total cost of running a data center can be chalked up to stranded power and cooling capacity, according to Wade. Modeling can help data centers avoid getting into the situation of not making optimal use of space, power, and cooling capacity.

For example, a data center may want to install a server in a rack, and the DCIM may indicate that there's space and power to put a new piece of equipment in. But is there enough cooling available in the rack? Without CFD, if the equipment is installed, the cooling problem may not come to light until after the new piece of equipment is installed, Wade says. And taking the server out after it has been installed might be problematic, because of fear of an outage. With CFD, however, it's possible to add new equipment to the model and learn whether enough cooling exists in that place to support the added power. If not, new equipment can be moved around (virtually) until the right location is found.

Predictive modeling also allows the data center facility manager to be sure about energy savings, Wade says, and can predict what will happen in a failure scenario, if the cooling system breaks down. "You can predict how long the servers will stay up before they overheat and shut down," he says.




Contact FacilitiesNet Editorial Staff »

  posted on 1/22/2015   Article Use Policy

Comments