13 June, 2019
Over the years that I have been involved in the data center industry, I have read numerous articles on power and cooling best practices to solve critical problems from a compliance, efficiency and capacity standpoint.
In his recent blog “6 Data Center Design Best Practices,” Mark Gaydos touches on the six best practices for data center operators, one of which focuses on having a power and cooling strategy in place to maintain a highly efficient data center. The engineer in me appreciates the fact that best practices are available to ensure that problems are solved using tried-and-true methods. However, when it comes to optimizing power and cooling systems in a data center, I don’t quite agree that current best practices can solve the critical problems faced by today’s owner-operators.
Specifically, Gaydos suggests that organizations should use containment to manage airflow and cooling. The concept of controlling airflow using containment is nothing new, and experience has repeatedly demonstrated that blanket solutions don’t work. I have seen successful, as well as unsuccessful, installations of hot and cold aisle containments.
It's a difficult balance to strike. Under-pressurize the aisle and the servers will pull in air from leakages in and around the cabinets. Over-pressurize the aisle and you will be wasting air and reducing the cooling efficiency.
Figure 1. Recirculation inside a server can be difficult to manage
To balance the required airflow to the supplied airflow effectively, you must ensure you are matching the air supply to demand for each IT equipment installed. Add cooling controls to the mix and now you have a complex system that not only needs to match the demand from existing load but can also match that of future loads.
Mark also alluded to having enough power and cooling to keep up with future changes. This is correct at the macro level since you need to ensure you match facility demands to meet business demands. However, at the micro level, this is a lot more difficult when it comes to cooling.
In my blog “Does your business have the capacity to change?” I demonstrate that, for the same total cooling, two layouts can yield different results.
Figure 2. Two layouts can produce different results despite the same total cooling
In my example, one layout works and the other doesn’t, demonstrating the complexity of managing cooling at a micro level. Both examples indicate that merely adopting "best practices" doesn’t guarantee improvement in data center performance without CFD simulation. As such, CFD simulation must also be recognized as a best practice for designing and operating data centers.
With a data center digital twin, you can predict all possible changes to your data center and pick the ones that meet your objectives as well as your budget. I am not advocating dropping best practices, but isn’t it time that the data center industry adopts CFD simulation as a best practice?
Blog written by: Akhil Docca, Director of Marketing
Other Recent Posts
The Digital Twin for Hyperscale Data Centers
In a world where artificial intelligence and machine learning are the new frontiers, cloud applicat…
Model Server Virtualization in the Data Center Digital Twin
A recent study by Uptime Institute surveyed 250 C-level executives and data center/ IT managers at…