In the data center industry, Computational Fluid Dynamics (CFD), and particularly its application to modelling of data hall air distribution, tends to polarise opinions. Its detractors highlight its limitations to represent the complexity of reality whilst its proponents advocate its use for investigation.

Interestingly, both sides agree that the modelling is only as good as the information fed into it and the modeller training and experience. I, a proponent of CFD modelling, see CFD modelling software as a tool, which can be and is applied in the modelling of air distribution in data center white space.

CFD modelling
CFD modelling – RED Engineering

The aim of this opinion is to share what I hope is a balanced view on applying CFD modelling to bring value to project through building confidence in the design through assessment of quantified outcomes and education, rather than just for modelling sake.

First, let’s have a look at the motivations. The following data center design guidelines nominate CFD modelling in the context of data hall white space:

  • CFD nominated as ‘a feasible alternative for testing various design layouts and configurations’ [ASHRAE Design Considerations for Datacom Equipment centers, Second Edition];
  • ‘For best results utilize a combination of […] and a CFD modelling system for power and cooling capabilities’ [ANSI/BICSI 002-2014 Data Center Design and Implementation Best Practices];
  • ‘Consider computational fluid dynamics (CFD) models for large data centers to optimize location of telecommunications pathways, air conditioning equipment, equipment enclosures, air return, air vents and ventilated tiles’ [TIA Standard TIA-942-A Telecommunications Infrastructure Standard for Data Centers, August 2012];
  • ‘Air management is an energy enabler, […]. One management solution is to use modelling and simulation, such as data-center infrastructure management (DCIM) or computational fluid dynamics (CFD) tools to plan data center changes.’ [CIBSE KS18: Data centers: An Introduction to Concepts & Design].

These recommendations are powerful drivers to promote white space CFD modelling as best practice and hint at benefits to be gained from applying CFD modelling to data center white spaces such as testing and optimising designs. But these recommendations also raise the following key questions: does undertaking a CFD analysis yield these benefits by itself and what are those benefits really?

To answer these questions, one needs to consider is CFD modelling itself. A CFD modelling software is a tool that solves a set of equations that can predict airflow and temperature distribution. This tool is to be tailored to the application by the CFD practitioner.

Guidance on the application of CFD modelling for environmental modelling can be found in ASHRAE RP-1133 “How to verify, validate and report indoor environment modelling CFD analysis” or CIBSE AM11 “Building Performance Modelling.” While these documents provide guidance on the application of CFD, this guidance is not specific to the data hall airflow modelling

The ASHRAE Research Project RP-1675 “Guidance for CFD Modelling of Data Centers” is expected to yield data center specific guidelines and benchmarks that will benefit such application. Also, the use of a data center specific CFD software clearly assists with capturing key modelling aspects, such as leakage path or behaviour of floor tiles. Is this sufficient to capture the value of CFD modelling of data center white space?

In my opinion, the key point in unlocking the value of CFD modelling, when not undertaken for the sole purpose of having CFD report, is the understanding of the project by the CFD practitioner and his or her capacity in translating this understanding into the right model. In this area, I favoured an approach inspired from the Ten Iterative steps in development and evaluation of environmental models, A.J. Jakeman, R.A. Letcher, J.P. Norton, Environmental Modelling & Software 21 (2006) 602-614. Prior to undertaking the modelling itself, the following are identified and shared with the project team:

1. Modelling purpose;

2. Performance metrics and acceptance criteria; and

3. Key characteristics of the model.

In the context of data hall design, the CFD modelling purpose would typically be to confirm the adequacy of the cooling system to deliver the supply air to the IT equipment. The performance metrics could be based on the IT equipment inlet air temperature, temperature at cabinet front rack or temperature sensor location nominated in the SLA. One of the key characteristics of such a model is to ensure that it captures the elements used in the design of the cooling system, and hence is aligned with the basis of design.

Whilst this does not necessarily improve the modelling by itself, it ensures that the person doing the modelling has a thorough understanding of the intent of the model and is in a position to capture this intent when developing the model, rather than just ‘run a CFD model’. This also allows the CFD modelling study to become a discussion and learning exercise on some of the key aspects of white space airflow management: balancing cooling and airflow demand and supply, air leakage management, and control strategy. This then becomes a common ground on which to discuss optimisation strategies, such as positioning of cooling units and baffles for airflow management.

In many cases, another aspect is also discussed: scenarios to be investigated. This discussion centers around questions such as “is this a worst-case scenario” or “how likely is this failure scenario”. While the modelling is used to quantify the impact and hence its severity, the assessment of the probability is beyond the scope of the modelling and involvement of the wider team and client is required to clearly define the risk level. This is the step where the decision on whether to analyse the white space at part load and/or under fully populated white space is taken, based on the information available at the time.

In my opinion, the value of a data hall white space CFD modelling study lays in how it builds confidence in the proposed design. In my experience this is achieved through education on the modelling, including not only its advantages but also its limitations, and interaction with the design team and client to ensure ‘buy-in’ throughout rather than a “tick-the-box” exercise. When the CFD modelling tool is approach this way, it offers many unique insights into airflow and cooling features that are difficult to be gained otherwise.

Doctor Julien de Charentenay is head of Building Physics at building design consultancy RED Europe