Data centres: Risk Mitigation through Strategic CFD analysis

By Harshad Joshi on Monday 26 November 2018

Data centres are the cornerstone of high value mission critical infrastructure in today’s world of ‘Big Data’.

A data centre is a large space loaded with computer servers. The large amount of heat produced results in high space cooling requirements. This high-density heat load also means the right amounts of cooling airflow needs to reach all the computer equipment. Any starvation in the cooling air supply to the equipment risks failure due to overheating, or in extreme cases, could even lead to fire. This all means that efficient distribution of cooling air is paramount in safe and effective data centre operation.

The best way to analyse and predict the airflow design risk is by performing a Computational Fluid Dynamics (CFD) simulation of the airflow within in a data centre. CFD can model and simulate all aspects of equipment arrangement including:

  1. Perforated floor/ceiling tiles
  2. Hot/cold aisle containment
  3. Diversity in of heat load distribution on each server
  4. Underfloor/overhead plenum
  5. Cold air lobby

In order to mitigate risks, a range of modelling simulations should be performed to cover the range of anticipated operational modes of the data centre, including the following four main analysis types:

A. Design Airflow Analysis

This baseline airflow analysis of the data hall includes the design heat gains in each sever of the data hall with the cooling equipment operating at its optimum level. This type of analysis provides a snapshot of the typical operational conditions expected to exist. This has the potential to identify a range of potential issues including:

  1. Damper Control for Supply Grilles: When all the supply plenum grilles are set at constant opening the air distribution can become uneven due to the pressure distribution in the supply plenum. CFD can help to identify the right opening area/control for each grille to ensure the most effective approach in achieving equal/requisite airflow through each grille.
  2. Server Configuration: Identify high powered servers within the data hall and quantify where high cold airflow is supplied.
  3. Hot Spots: Preventing cold air picking up heat before entering the servers.

Figure 1: Design Airflow Analysis

B. Day One Analysis

Many data centres built today are multi-tenant. Therefore, the data centre may be designed to meet a certain heat load demand but will not reach this level until many days, weeks or months into the actual operation. So further analysis can be performed where several units can be set as ‘empty’ with the analysis identifying the most economical way to cool the units that are operational. This works in predicting the operation and running costs during the initial stages of operation.


Figure 2: Day 1 analysis

C. Equipment Switchover Simulation

Data centres are designed with redundant systems so that the operations can continue unhindered under all foreseeable conditions. One of the many potential failures that could occur is cooling units supplying cold air into the data hall. A cooling unit failure would result in heat build-up due to high-powered servers and in some cases a potential fire risk. This risk can even exist in the time it takes the ‘redundant’ secondary cooling unit to come online and start providing cool air. CFD simulations will establish the outcomes of these failure scenarios. The analysis can nominate which cooling unit should act as redundant so that if failure occurs then a hot spot will not arise which can lead to equipment damage.

Figure 3: Equipment switchover analysis

D. Leakage Analysis

As with the Day One Analysis, multi-tenant data centres also face the challenge of different server use as the full server rack may at times remain unutilised. Blanking plates would typically be used to block airflow at empty rack slots. CFD analysis can be performed to determine which racks locations are best to leave empty. This would help determine the effect from cold air leakage into the hot aisle or hot air into cold aisle when containment features are used.

Figure 4: Leakage analysis

CFD modelling is a relatively low cost way to help plan for a range of different contingencies whereby design teams can perform as many different scenario analysis as they like in advance of construction or at a later stage in order to minimise the risks and mitigate potential failures in such a mission critical environment.

IES Consulting’s CFD experts use a combination of IESVE and our partner CFD software package derived from the OpenFOAM® libraries. These allow us to perform a vast range of complex analyses in the most challenging environmental conditions including data centres, cold storage, cleanrooms, external pollutant dispersal, and pedestrian comfort in any climate zone.

To find out more about the benefits of this service, and how IES Consulting can help you contact us via http://www.iesve.com/consulting/contact