Jim VanGilder, a principal engineer with American Power Conversion Corp.'s (APC) data center cooling simulation...
group, recently gave a presentation at the American Society of Heating, Refrigerating and Air-Conditioning Engineers' (ASHRAE) annual conference in Dallas regarding a new metric to measure airflow efficiency. We caught up with VanGilder to ask him what it's all about.
Can you explain the capture index?
Jim VanGilder: It's a metric expressed as a percentage from zero to 100 for each rack in a data center. The value of 100% implies that all the airflow, either entering or leaving the rack, is following a desirable path.
What do you mean by 'desirable path'?'
VanGilder: It depends on the application, but if you think of a common raised-floor design, you would like the airflow to come from the perforated tiles and then go into the racks. You wouldn't want air to the racks to come from the room. As for leaving the rack, the desirable path would be when you have local return vents or local coolers that capture the air coming from the racks.
Why are there two versions of the capture index, the hot aisle and cold aisle?
VanGilder: The capture index may be used to assess either delivery of cooling to the rack -- that's the cold-aisle version. Or it can be used to measure the scavenging or capturing of hot air from the rack, and that's the hot-aisle version.
The hot-aisle capture index is generally used if the design includes local coolers, like APC's InRow cooling products, or some other type of local return vent. This is because if you're capturing the hot air locally, you can design your data center around a room-neutral strategy, the goal of which is to prevent any hot air exhaust (from equipment) from escaping into the data center room environment. The benefit is the vast majority of air in the data center is maintained at some desirable room temperature. Since all the IT equipment is going to be provided with consistent room temperature air, regardless of where it comes from, you don't have to worry about the delivery of the cooling air to the rack. All IT equipment, whether at the bottom or the top of the rack, will see uniformly conditioned air.
In a typical raised-floor facility, however, the hot exhaust air isn't captured locally, but typically travels all the way back to CRAC units, which are located across the room. So you have to focus on the delivery of the cold air.
So you use the hot-aisle capture index as the preferred method if you can, if you have local cooling methods. Use the cold-aisle capture index if you don't.
There are already a lot of metrics out there for data centers. What differentiates this one, and why is it important?
VanGilder: There are other metrics, and each of them serves a valuable function. The main differentiator is that this is the only metric based on airflow patterns. Others are based on temperature.
The temperatures don't tell the whole story when it comes to cooling performance. Imagine a group of racks in a typical raised-floor environment, and these racks are clustered around a single cold aisle with perforated tiles. Now imagine a really horrible airflow pattern where some racks are drawing exclusively from the ends of the aisle or from recirculated air from the other side of the racks, and not the perforated tiles. If the surrounding room temperature is low enough, and there's enough air movement around, the rack inlet temperatures might be OK. But intuitively, this is not the kind of data center we want to design because as the surrounding room temperature changes, the rack inlet temperatures will change.
However, if we had computed the capture indices in the beginning, we would have known that the rack inlet temperatures were OK, but the capture indices were not. That could compromise cooling performance down the road.
Will having a good capture index lead to lower cooling costs?
VanGilder: I think that's a good point, but I'm not sure it would hold in all cases. It's application specific. I think it's a likely scenario because the index is a measure of whether the air is going where you want it to go.
How can data centers measure their own capture index?
VanGilder: The capture index is typically computed in the design process as part of a CFD [computational fluid dynamics] analysis or using some other dedicated tool. The capture index can help you find an optimum range of equipment and discover where potential airflow problems may exist through failure scenario investigations.
Physically measuring the capture index in a working data center is difficult. You could do it with a smoke test, but it's something that is best predicted numerically. But those who want to see what the problems are, if they were to do a CFD analysis or use APC's InfraStruXure Designer, it would provide some good clues as to what the underlying problems are. It would hasten the way toward a solution.
How did the capture index come about?
Van Gilder: My team is responsible for developing cooling calculators for InfraStruXure Designer software. The software is used by APC engineers and channel partners to provide data center design solutions for our clients. We also have a group called the APC Science Center, which is responsible for vast numbers of white papers on various topics, so some of the work by the APC Science Center fed into the work my team has done.