Benefits of CRAC Supply Air Temperature Control
Controlling the temperature in a data centre is essential for maximizing uptime and efficiency, but what is the best way to achieve this? While data centre architecture has progressed quickly to separate hot and cold air, with hot aisle/cold aisle designs becoming industry standard, most air handler temperature control tactics are still based on decades-old mainframe and comms room ideas. By adjusting the temperature of the air supplied by the different cooling units, return air temperature management seeks to achieve a constant ambient temperature within the room. This frequently results in a substantial change in the temperature of the air delivered to the IT equipment, a change that is beyond the facility or IT manager's control. Controlling the temperature on the supply side of the air handler instead of the return offers the facility management more control and makes the temperatures provided to servers more predictable. This predictability allows for higher water temperatures and lower operating costs, as well as capital cost reductions during the construction of new facilities.
In the data centre environment, the need of airflow management and precise temperature control has become increasingly apparent. With new server designs and power densities, old design solutions that simply threw the necessary kW cooling into the data centre space and regulated the ambient temperature were unable to cope. It's easy to forget that the data centre cooling system's sole purpose is to give air of an acceptable temperature and humidity to the IT equipment (at their inlets), not to promote occupant comfort or maintain general room air temperatures. Given this, it's clear that the return air temperature isn't a good indicator of the temperature of the IT equipment inlet, and it seems like expressly managing the temperature of the air supplied by the system would be preferable. The facilities manager gives up practically all control over the temperature of the air supplied to the IT equipment when adopting a return air control technique. Switching to supply air control will return control to the facility manager, but how will this affect the data centre’s operation? Many facility managers say that having the units regulate the ambient temperature by varying their supply temperatures allows them to respond more quickly to changing loads within the data centre.
Money Saving Potential
There is sufficient headroom between the chiller set point and the cooling unit set point thanks to the supply control approach and the ensuing stable, predictable temperature in the floor void. By boosting the chilled water temperature, this opens the door to potential energy savings. The only step made to accomplish the temperature increases and cost savings was to modify the control system. There is still a lot of IT equipment that is hotter than the fixed temperature of 18°C. It is feasible to achieve further increases in supply temperature by taking action to lower this by making other design changes, such as partial enclosure of the cold aisle. Indeed, in a well-designed data centre with supply control, all server inlets should be within a degree or two of the control temperature, and supply temperature set points of up to 22°C or higher can be attained, resulting in significant cost savings.
Are there any other knock-on effects?
Switching to supply side control has proved to have no negative impact on servers, but are there any other drawbacks that would prevent the switch? It is often assumed that increasing the temperature in a data centre will reduce its resilience. However, a data centre is only as resilient as its least resilient equipment, which is 6.5°C below thermal shutdown in the return control approach. Controlling the supply temperature to 18°C has lowered this by half a degree to 6°C, although this is due to the temperature chosen rather than the control mechanism. Rack intake temperatures have been found to have a direct link with supply air temperature, therefore if a control temperature of 17°C had been used, a half-degree gain in resilience would have been noted, as well as the same cost savings. As fan speeds increase and/or components become less efficient, rising temperatures can increase the amount of power servers’ demand. Above 22°C, fan speeds tend to increase, and component inefficiencies become substantial. The number of servers over any of these temperatures has not increased much because of changing control measures, therefore this effect is likely to be minor. Finally, working circumstances in the data centre will have changed, though not necessarily for the worse. Hot aisle temperatures have risen a few degrees from 22-23°C to 25-26°C, which is still tolerable; cold aisle temperatures have risen significantly from 13-15°C to a more pleasant 18°C.
It has been shown that moving to a supply temperature management technique offers considerable benefits. Cooling units are less susceptible to sensor placement and inter-unit crosstalk, allowing the facility manager to have more control over the temperature provided to the IT equipment. As the facility grows, a supply control method can keep up with the increased demand while keeping server temperatures within a narrower, more manageable band. Finally, adjusting the supply temperature ensures consistent headroom between chilled water and supply air temperatures. Closing this gap by increasing the temperature of chilled water will result in energy savings and lower operating expenses. Because the performance and potential savings will differ from one facility to the next, the benefits need to be determined on an individual basis.