In this edition of Voices of the Industry, Eric Jensen, Vice President & General Manager at Data Aire discusses how data centers can achieve high density cooling at scale.
We are experiencing a shift in the data center space and this is a moment for you to reflectively and respectively ask what does that transition look like for me? Am I trying to improve operations and improve efficiency, and does this transition need to be a nightmare? There are opposing forces at play in the data center industry today. It seems that while big facilities are getting bigger, there are also architectures that are trying to shrink footprints. So, as a result, densities are increasing. What we hear a lot about is 50, 70, even 100 kilowatts per rack because high performance compute is fun to talk about. However, I find that for the near future, it’s less about the highest densities available and more about high[er] densities.
According to AFCOM’s State of the Data Center Industry report, and a 2021 report from the Uptime Institute, densities are now averaging 6-12kW/rack. These increases, along with designers’ and operators’ desire to plan for the future is creating a gap between the design load, and the actual load.
To manage this gap requires scalable solutions. Nobody wants to get caught unawares a few years down the road. So, it’s understandable to want to design for 12, 15 kw per rack. But the reality of many operators is still in that 6, 8, 10, 12 range and so how do you reconcile that? And that range is happening for a number of different reasons. Perhaps it’s because of the scaling of deployment over time or it’s because of the seasonality inherent of a tenant’s business.
I hear conversations centered on filling the bucket of spaces already available rather than building massive new facilities. The other discussions I tend to hear revolve around density efficiency — being able to support an economy of scale in a much more sustainable manner.
Higher Density at Scale
You still have to satisfy that need for scalability and those higher densities of load is still achievable in the same kinds of traditional ways. Whether you’re talking about chilled water for some facilities or DX solutions (refrigerant based solutions), for other types of facilities, both can achieve these higher densities in the traditional perimeter cooling methodologies without the need for completely rethinking the way that you manage your data center and the load coming from those servers.
And so, whether if chilled water solutions are doing it today because those systems are getting much larger at the cooling unit level, that’s satisfied simply by higher CFM per ton. Greater airflow delivery per ton of cooling is achievable without the need to dramatically alter the way you operate your data center, which is important nowadays because, to reiterate, every operator is in transition mode — not only their IT architecture, but also their power side, and their cooling infrastructure.
Those transitions are more manageable when your cooling systems are engineered-to-order. For chilled water solutions, multi-fan arrays are very scalable. You can scale down from 25 to 100 percent for the delivery, depending on whether you are trying to scale over the life of the buildout or you’re scaling again back to the seasonality of the business for whomever is the IT consumer.
And if it’s DX solutions, that’s achievable from a good, better, best scenario. ‘Good’ did the job back in the two to four kilowatt per rack days. But now, variable speed technologies are out there, and they can scale all the way from 25 to 100 percent. And they do so efficiently.
What we’re seeing at Data Aire is a lot of systems designed at the facility level leaning towards dual cooling. And so, dual cooling affords the redundancy of the infrastructure. In the data center world, we like to see redundancy. But it also introduces the opportunity for economization.