Many IT managers look only at server heat loads when determining the cooling needs of their data centers. However, there are many other components that contribute to the heat calculation. Networking gear, storage devices, and other IT equipment generate heat, which can be calculated based upon the amount of power consumed.
The lighting in the facility adds to the heat, depending upon the types of fixtures, wattage, and hours of use. Staff in the data center also generate heat, to the tune of about 550 BTUs per person, per hour.
Uninterruptible power supply (UPS) units are commonly overlooked as sources of heat. However, a UPS will sometimes give off more heat than the server it supports. This can make a big difference in terms of heat load, particularly in a small data center.
In a typical data center, the relative contribution of various components to the total thermal output looks something like this:
UPS units generate heat even when they’re not supporting the IT load. Heat is output via the power loss of the electrical circuits within the UPS.
The thermal output of a UPS includes both the fixed loss and the loss proportionate to its operating power. The calculation looks something like this, assuming that the unit is 95 percent efficient:
(0.05 x power system rating) + (0.05 x total power of the IT load) = heat load in watts
The spec sheet for the UPS will normally provide the heat output of the unit at various levels of IT load. Note that UPS units are most efficient when they are operating at (or near) 100 percent of their rated capacity. Therefore, the specifications may provide a range of efficiency ratings depending upon the IT equipment load. Efficiency may drop to 90 percent if the UPS unit is operating at 25 percent of capacity.
Lithium-ion batteries are more efficient than valve-regulated lead-acid (VRLA) batteries. UPS units that use lithium-ion batteries operate at 95 percent or greater efficiency, which means that they dissipate less heat.
Some manufacturers say their UPS units are 99 percent efficient because they provide an “ECO” or “green” mode of operation. In ECO mode, the inverter remains offline until needed, saving electricity and therefore reducing heat.
The problem is that it takes time for the UPS to detect a power failure, turn on the inverter, and supply clean power to the IT equipment. That process can take up to 16 milliseconds. While the typical server may not be adversely affected, this delay can negatively impact other IT equipment and components in the data center.
Line-interactive UPS units provide up to 98 percent efficiency without the risks of ECO mode. Because the inverter remains online, the load transfer time is less than four milliseconds.