High-Efficiency Cooling Systems for Data Centers

Start with current and planned IT load, then add UPS and power-distribution losses, lighting, building-envelope gains and auxiliary-area loads. Assess rack density and the share of liquid cooling separately. Do not add standby chiller capacity to the cooling load; determine the number of duty and standby units separately.

Data centre load changes over time, and a new facility may operate below its design capacity for years. Assess part-load measures such as IPLV or SEER, minimum capacity, compressor staging and pump and fan energy. An annual energy analysis based on the facility’s hourly load profile provides the most useful comparison.

Yes, if the IT equipment, CRAH capacity and rack-inlet conditions allow it. ASHRAE TC 9.9 recommends an 18–27 °C server-inlet range for A1 equipment, allowing 18/24 °C chilled water or higher in many facilities. Higher water temperature can improve chiller efficiency and extend free-cooling hours, but any increase in CRAH fan energy must also be assessed.

A common threshold is 20–30 kW per rack; above it, the airflow and fan energy needed for air cooling can become uneconomical. Air cooling with aisle containment may be sufficient up to about 10–15 kW per rack, while GPU clusters at 80 kW per rack or more require direct-to-chip liquid cooling.

The hours available for full free cooling depend on the site climate and water temperatures. Partial free cooling may be used when outdoor air is not cool enough to meet the whole load. Calculate operating hours from hourly weather data, dry-cooler approach temperature, glycol concentration and the actual facility load.

Water-cooled systems can be highly efficient under suitable conditions, but need cooling-tower water, pumps and water treatment, which increases WUE. Air-cooled systems avoid water consumption and open-circuit maintenance. Compare capacity, climate, water availability, energy cost, space and maintenance resources.

N is the equipment capacity needed to meet the design load. N+1 adds one standby unit; 2N duplicates the full system so either side can carry the whole load. Also examine single points of failure in pumps, heat exchangers, controls, valves and power supplies.

The source design calls for at least N+1 redundancy across cooling components so any unit can be maintained without affecting the IT load. For Tier IV, it calls for two independent cooling systems (2N), each with a separate power supply.

Possibly, but first check facility-water temperature, available capacity, CDU heat-exchanger approach, rack flow and water quality. Add the new liquid-cooling load to the existing air-cooling load and verify that the pipes and pumps can deliver the required flow.

The source design targets PUE below 1.3 for new facilities and considers 1.15–1.25 achievable in climates with strong free-cooling potential. In an existing facility, PUE above 1.6 can indicate substantial room for cooling-system improvements.

No. PUE covers cooling, power distribution, lighting and other auxiliary energy use across the facility. Airflow management, pump and fan control, UPS losses and IT equipment efficiency all matter alongside chiller performance.