Main author

BSRIA Institute / association Website
Last edited 27 Sep 2016

Data centre cooling

This article originally appeared in BSRIA’s Delta T magazine January 2011. It has been posted here by --BSRIA.


BSRIA’s head of computer and physical modelling, William Booth, provides the answers to technical questions on data centre cooling posed by BSRIA members.

Contents

[edit] Optimising cooling

Q. In what ways can cooling systems for data centres be optimised to achieve the best performance?

Optimisation of data centre cooling can be viewed on the micro (component) level but also on the macro level to ensure that the data centre operates efficiently, reliably and effectively.

Measurement schemes, such as Energy Efficiency Rating (EER), enable operators to move towards optimal performance. Another driver for efficient operation can be the knock-on benefits to company profitability. Alternatively, as most data centre sites will have a limit of maximum available electrical power, optimisation will allow the operator to maximise the efficiency of a system while maintaining IT functions.

The principles of data centre cooling are straightforward. The main challenges are the implementation of the design and the operation of the data centre cooling and IT equipment. The quality and quantity of cooling are important. It calls for attention to detail and good understanding, co-ordination and management at the micro and macro levels.

[edit] Floor grilles

Q. To minimise the number of floor grilles we are using a chessboard floor pattern. We have a problem with cooling. Why?

Pictured below is an infrared image (warmer temperatures are red, cooler are blue) of two sections of the cold aisle in a data centre. In the foreground, the grilles are arranged in a continuous line across the front of the grilles. In the background, the grilles have been arranged in a chess board arrangement. This shows clearly the effect of the two different arrangements.

Thermal image of floor grilles in data centre.jpg

Image: Infrared image showing two sections of a cold aisle in a data centre.

The cabinets in the foreground show a uniform distribution of cold air across the front with the hot air being kept at the top where it can be extracted. The cabinets in the background show a non-uniform distribution of cold air.

The hot air is re-circulating and coming out the front of the cabinet at low level, while some of the cold air is by-passing the front of the cabinet and punching through the hot air at high level to cool the ceiling.

This graphically answers the question that if you use a chessboard pattern for the floor grilles, the effectiveness of the cooling is compromised. More generally, uniformity in cooling requires uniformity in air distribution.

[edit] Data centre operation

Q. How can you troubleshoot the operation of a data centre?

Troubleshooting the operation of a data centre requires a logical approach. This may appear a daunting task for a live data centre, but successful treatment can only come from a correct diagnosis.

The precise approach taken will depend on whether a step-change in performance has been noted, newly-installed equipment is not being cooled as expected, a non-resilient response to a computer room air conditioning (CRAC) unit or chiller unit failure has occurred, or there is a gradual fall in performance. Perhaps the operator wishes to take a proactive approach to performance and efficiency levels.

BSRIA’s BG 5/2003 Cooling Solutions for IT provides more guidance, but the three golden rules of thumb are:

  • Think of warm air as a pollutant
  • Don’t cut holes in the floor
  • Don’t over- promise or over-demand.

Assistance may be required to test prototype cabinet systems, perform validation by testing mock-ups, of for on-site trouble-shooting services.

[edit] Validating performance

Q. How can I validate performance at commissioning?

The performance of a data centre depends on supplying the correct cooling to all the installed equipment. A range of acceptable conditions for the cooling supply will be set for the commissioning tests to keep the equipment running efficiently without interruption. Supply conditions need to be measured to establish if they are correct (particularly temperature and humidity), and validated for all foreseeable eventualities, such as failure modes, at full and part-load operation.

A versatile testing system is needed to validate the large range of layouts of data centres and the different types of installed cabinets used. An independent validation check of the building management system (BMS) and associated control and monitoring strategy allows the operator to go into the operational phase with greater confidence.

[edit] Key temperatures

Q. How and where can you measure the key temperatures in a data centre?

Maintaining an environment appropriate for the IT equipment, coupled with an uninterruptible power supply within a secure space, are the primary purposes of a data centre. Whether for monitoring for compliance or control purposes, temperature measurements must be appropriate, accurate and averaged. Temperatures can vary widely depending on where readings are taken. This is important to know, as taking temperature readings in the wrong spot can lead down the wrong path, resulting in wasted time and money and potential damage to business-critical equipment.

[edit] Find out more

[edit] Related articles on Designing Buildings Wiki

[edit] External references