With advances in technology, servers are becoming faster and more powerful. They are also generating significantly more heat. Server equipment is designed to operate within a specific temperature range and staying within this range will ensure reliable operation and extend the life of the equipment. On the other hand, exceeding these limits could lead to catastrophic server failure which can cost companies in equipment replacement, lost time, lost data and lost business.

Often, cooling systems are over-sized to provide a “factor of safety” against catastrophic failure. However, more isn’t always better because an over-sized cooling system wastes energy and operates inefficiently. The additional cooling capacity will bring the server room temperature down quickly. The compressor associated with the cooling unit will cycle on and off rapidly over the course of a day to maintain the server room temperature set-point. This is known as short-cycling, which will reduce the life expectancy of the compressor. In addition, drastic temperature swings can be damaging to the server equipment.

To determine the size of the cooling system required, all heat sources must be identified. Anyone that will be working in the space needs to be accounted for along with heat generated from lights. Most data centers are located in the interior of the building; however, if there are any outside exposures (walls, roof, windows) these will need to be taken into account. The largest and most difficult heat source to calculate is the equipment load.

Typically, engineers will estimate the heat load generated by computing the power consumption of each piece of equipment. With nameplate data for the amperage and voltage, the wattage can be determined using Ohm’s Law (Watts = Volts x Amps). However, this usually results in loads 30-40% higher than actually present. There are several contributing factors to these inaccuracies:

1. Manufacturers list the highest rated amperage that the equipment is capable of drawing. This is based on full load operation of the equipment which is not expected in real life for extended periods of time.
2. The amperage listed on the nameplate can be rounded up. For example, 3.6 amps might be shown as 4 amps.
3. The nameplate amperage does not occur at the nameplate voltage. Power consumption by the equipment remains constant. According to Ohm’s Law, as the current increases the voltage must decrease to maintain the same wattage. Therefore, the nameplate amperage is based on this reduced voltage which is unknown.

A more accurate method is to obtain heat output ratings of the equipment from the hardware manufacturers whenever possible. It is important to model the actual equipment operation. For example, any equipment that is provided for redundancy will not operate simultaneously to its counterpart and should not be added to the expected heat load. Additionally, it is extremely rare for equipment to operate at full load capacity for extended periods of time. If the hardware is currently installed and operating it is sometimes possible to measure the actual heat load. Many racks have the ability to measure the power consumption of associated equipment. It is important to average multiple readings over the course of the day to ensure that you are getting an accurate representation of equipment utilization. Finally, accounting for any future growth up front should be considered as there is a potential to save considerable time and money.