A comparison of air cooling and liquid cooling is only meaningful when both designs include the same energy boundary.
Account for the path from IT heat generation through distribution and facility heat rejection, including relevant fans, cooling distribution units, pumps, air handlers, and heat-rejection equipment. PUE and heat-reuse metrics can help, but neither replaces a clear record of what was measured.
Data center deep-dive series
Start with a common energy boundary
The question is not simply whether air cooling or liquid cooling uses less power. It is whether the two designs are being measured across the same part of the system. Comparing liquid-cooled server fan power with an air-cooled facility’s air-handler and chiller power would compare different boundaries rather than different cooling approaches.
For a facility-level comparison, use a boundary that follows IT heat from the equipment through the cooling path to facility heat rejection. Depending on the design, this can include server and switch fans, air-handling equipment, cooling distribution units (CDUs), IT-side and facility-side pumps, chilled-water equipment, economizers, chillers, dry coolers, or cooling towers.
Power Usage Effectiveness (PUE) is the ratio of a data center facility’s total annual energy to the annual energy drawn by all IT equipment. It can indicate the burden of supporting infrastructure, but it does not identify which individual component—such as server fans, a CDU, pumps, or a chiller—caused a difference between two designs.
U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design
The first heat-transfer path is different
Traditional air cooling transfers heat from IT equipment to room air and then conditions that air through the data center cooling system. Rack-inlet conditions, exhaust-air recirculation, and bypass paths through open rack spaces, cable openings, or other gaps can affect the cooling system’s performance and auxiliary energy use.
Direct liquid cooling transfers waste heat to a fluid at or near the point where it is generated, rather than first transferring it to room air. The U.S. Department of Energy describes this approach as capable of supporting higher heat densities than traditional air cooling.
“Liquid cooled” does not necessarily mean that liquid removes all IT heat. Some implementations are hybrid: liquid captures most of the heat load while conventional air cooling removes the remainder. Other approaches can capture practically all heat without using fans. A useful comparison therefore records both the heat handled by the liquid path and the residual air-cooling load.
Do not exclude the CDU and facility loop from liquid cooling
Many liquid-cooling approaches use a cooling distribution unit. A CDU interfaces between the facility cooling loop and the IT cooling loop, supplying liquid at conditions appropriate for the IT equipment, including temperature, pressure, and fluid chemistry. Measuring only reduced server-fan energy while excluding the CDU and the connected facility cooling path leaves part of the heat-transfer system outside the comparison.
Liquid cooling can potentially reduce the required capacity and power of fans. Under suitable design conditions, it may also work with medium-temperature chilled-water supply and water-side economizers. Higher liquid supply conditions may make atmospheric heat rejection with dry coolers feasible in some designs.
These are conditional design opportunities, not a universal efficiency result for every liquid-cooling implementation. Supply conditions, heat-rejection configuration, residual air load, and controls all affect the result.
Record the comparison in three layers
Separating measurements into three layers makes it easier to see where energy is being used rather than combining unlike loads in one headline figure.
IT equipment
- Record
- Server fans, switch fans, total IT energy
- Question answered
- How did the cooling approach affect power within IT equipment?
Rack and distribution
- Record
- CDU, rack or in-row equipment, IT-side pumps, residual air-cooling load
- Question answered
- How much heat remains on the air path, and what does the liquid-distribution path consume?
Facility heat rejection
- Record
- Air handlers, facility-side pumps, chilled-water equipment, economizers, chillers, dry coolers, or cooling towers
- Question answered
- What auxiliary energy is needed to move heat to the facility heat-rejection boundary?
Also document the IT load, measurement period, relevant inlet-air or liquid-supply conditions, and reliability or redundancy conditions for each design. Measured values should be trended continuously and retained for at least one year when annual-energy comparisons are intended. If the comparison uses only a different season or load interval for each design, state that limitation with the result.
Treat heat reuse as a separate accounting decision
If waste heat is delivered for use outside the data center, do not automatically combine that benefit with cooling-energy performance. Energy Reuse Effectiveness (ERE) is defined as total annual facility energy minus reused energy, divided by IT equipment energy. The reused energy is recognized outside the data center control volume.
A heat-reuse comparison should therefore state where the heat was metered, what external use received it, and where reuse was recognized. Reporting cooling energy without heat reuse alongside a value that includes heat-reuse credit as though they were the same efficiency measure makes the outcome difficult to interpret.
The more reliable framing is not “which cooling method is always more efficient?” It is: under the same IT load and operating conditions, what auxiliary energy is required from the point of heat generation through heat rejection or explicitly defined reuse?
No comments:
Post a Comment