On its face, row-based data center cooling sounds like it’s all about sending cool air to wherever it’s needed in order to keep IT equipment at optimum temperatures. But, as it turns out, that’s not really what it’s about at all: row-based cooling is actually about capturing hot air from IT equipment before it has a chance to mix with other air in the data center.
That’s the starting point you need to understand in order to dispel three misconceptions outlined in a new white paper by Schneider Electric, “How Row-based Data Center Cooling Works.” The paper imparts some valuable lessons that can help save any data center operator some money, by obviating the need to put cooling where it isn’t needed and buying equipment that’s not necessary.
The goal of any data center cooling system is to remove the heat added to the air by the IT equipment. As such, there are some best practices for effective hot air containment, including use of blanking panels and brush strips, proper row cooler placement, use of side air distribution units for switches and routers (which often employ side-to-side airflow) and use of air containment systems, among others. The paper also discusses a metric for measuring how much hot air is captured, known as the capture index.
But the meat of the paper is in dispelling three misconceptions about how row-based cooling technology works and how it’s deployed.
The first has to do with the idea that row coolers require the use of turning vanes, which direct cold air to the front of the racks. When properly implemented, a row-based cooling architecture requires no such thing. Why? Because if 100% of the hot IT exhaust air is captured and cooled, (before it has a chance to mix with surrounding air in the room), the remaining space in the room becomes a cold air supply plenum.
Misconception 2 is the idea that row coolers are required in every row. In fact, with a high hot air capture index, it doesn’t matter if all the row coolers are on one side of a two-row pod (or any placement combination). (A pod is a cluster of IT cabinets, power and cooling deployed as a unit.) The reason gets back to the point above, that the entire room is, in effect, a cold air supply plenum.
The final misconception is that row coolers can’t cool loads outside of their pod, meaning extra cooling capacity in one pod can’t help cool another pod. Here the paper shows a CFD analysis to prove that this is a misconception.
A key takeaway from this paper is that some companies, perhaps on the advice of their vendors, are taking steps that aren’t necessary if they have a properly implemented row-based cooling architecture. Turning vanes, coolers in every row, cooling sized for each pod – none of that may be necessary with a properly implemented row-based cooling system..
Check out white paper number 208, “How Row-based Data Center Cooling Works,” to learn more about the topic. It may save you some real money in your data center design and long-term operation.