Red engineering design responds...

In our award submission, the statement that our designs have cut data centre energy consumption by more than 30% was based on a benchmark of what we consider a standard data centre design. This would typically include:

• packaged air cooled chillers
• chilled waterflow and return temperatures of 6/12ºC
• 3-port valve arrangements on CRAC units
• fixed-speed fans on CRAC units – now that suppliers are starting to offer variable-speed drives as standard, they are finally getting installed on more data centres
• electric Vapac humidifiers in CRAC units
• air-conditioning to UPS rooms
• electrical heating coils to fresh air AHUs.

PUE is fast becoming the main benchmark for data centre energy performance. The strict definition of the PUE is as described in Mr Affleck’s letter, but ideally should be calculated on a design day as this is the only fair way of comparing all data centre designs on a reasonably equal basis. For example, by using the average annual PUE, a data centre in Scandinavia with free cooling could never be matched by a data centre in Barcelona.

The PUE should also consider all the powertrain losses and all the lights on and any small power and ventilation losses associated with the operation of the data centre. It would not include associated office space as this can vary substantially from site to site.

Red has recently completed Fujitsu’s London north data centre, which has 3300m2 net technical space at 900W/m2 average and will operate with a PUE of 1.41 based on the above criteria. It is good to see we are designing comfortably inside Hewlett-Packard’s “ideal” levels. The PUE is based on a design day of 37.9ºC dry bulb and 24ºC wet bulb, which is obviously rather onerous.

The chiller efficiencies increase greatly as the ambient temperatures move down from these figures and there is a significant amount of free cooling built into the design, which means the chillers often won’t run at all in the winter. This all results in a significantly lower average annual PUE of 1.28

Regarding the supposed blanket ban on open-circuit cooling towers in the event of an unidentified legionella outbreak, Paul McDermott of the Health and Safety Executive has confirmed that the normal procedures are as follows.

The authorities do not operate a “blanket ban” strategy on cooling tower systems in a legionella outbreak area and therefore will not shut down a cooling tower system if it is under control and being maintained and monitored in accordance with ACoP L8.

Prohibition notices are considered when a cooling tower is inspected by the authorities and found to be “not in control” from a bacteria point of view, and the inspector is of the opinion that the system poses a significant risk to human health.

Minor non-compliances with ACoP L8 guidelines may result in recommendations from the inspector. The continued operation of a cooling tower system is in the control of the operator and their appointed control and maintenance team. The authorities intervene when the system is found to be at risk and non-compliant with ACoP L8.

Dr McDermott stated: “HSE and local authority regulatory colleagues do not operate a ‘blanket ban’ on all evaporative cooling systems in the outbreak area. Inspectors will aim to visit all evaporative cooling systems in the area if there is epidemiological evidence that points to this type of risk system being responsible for the outbreak.

“The purpose of the inspections is to check the standards of operation, maintenance and control against those in ACoP L8. The use and type of formal enforcement taken would always be proportionate to the risks encountered and the issuing of improvement notices, pro-hibition notices and prosecution would only follow if there were substantial deviations from the ACoP.”