DCN October 2016 | Page 23

green IT It is calculated by dividing the total power consumption by the power consumption of just the IT equipment. Unfortunately, the PUE metric does not reward reductions in IT power. Since IT power is in the calculation’s denominator, even though IT energy consumption reduction is a direct and compounded benefit, it may actually increase the PUE value for a data centre. This should not discourage data centre and facility managers from pursuing these valuable ways to reduce energy and expenses. Minimise cooling fan power The next biggest energy savings opportunity is the power consumed by HVAC fans to move air in the data centre. IT equipment are virtually all air cooled devices. A cool air supply is needed to meet the demands of the IT equipment, and warmed air exhausted by that equipment needs to be removed. Airflow optimisation requires recognition that these two airflow volumes are actually the same. To minimise the amount of air that needs to be delivered, the mixing of cool supply air and warm exhaust air needs to be prevented. Data centre physical infrastructure deployment has evolved to do this. Hot aisle/cold aisle arrangements of IT equipment were an early step. Without airflow barriers however, air mixing at row ends and over the tops of rows still occurs, requiring as much as 20 per cent oversupply of cool air to offset this mixing. Efficient airflow management requires a containment methodology. Fully enclosed aisles, with endof-row doors and ceilings or other structures to contain the tops of the aisles are effective. These are known Cold aisle containment – with hot exhaust chimneys can mitigate cold air oversupply. as either cold aisle containment or hot aisle containment, depending on whether the IT inlet or exhaust rows are contained. Vertical exhaust ducts, or chimneys, on IT cabinets are another technique for containing the IT exhaust airflow. Most new data centres are deploying a containment system. Existing operations can consider ways that containment can be added, potentially with minimal disruption to operations. Deploying a physical air containment method is not the end of the optimisation process. Data centres are not homogeneous, static environments. Airflow demands are not uniform and they aren’t constant. It is likely that some rows, regions, or cabinets will have higher than average demand. True airflow optimisation requires information about the airflow requirements and adjustment of the flow volumes to meet the demand without over or undersupply. Commissioning for initial demands is a starting point, but environmental monitoring of air temperatures and pressures can be used to either recommission periodically or to automate airflow delivery to adjust dynamically as IT needs change. 23