OSDCTC.020: Measuring Rack Heat Removal With Flow and ΔT

Data center technician checking a tablet beside a liquid-cooled server rack, CDU piping and flow instrumentation; original neon-green BitcoinVersus.tech watermark on black at bottom left.

Elementary overview

OSDCTC.020 teaches a fundamental data-center cooling calculation: estimating how much heat a liquid loop removes from a rack using measured coolant flow and temperature rise. The goal is not to declare a cooling system healthy from one number. Instead, technicians learn what a measurement represents, where sensors must sit, and why apparent cooling capacity can differ from the rack’s electrical power. The lesson follows OSDCTC.019 on coolant and filter health.

HFM PHE, “Data Center Cooling EP.2 | The Hidden Trade-Off Between ΔT, Flow Rate & Pressure Drop” (June 2026). Manufacturer explanation of operating-point trade-offs in heat exchangers and CDUs.

The heat-transfer equation

For steady, single-phase liquid cooling, the relevant relationship is heat transfer = mass-flow rate × specific heat capacity × temperature rise. Use the coolant’s measured density and specific heat, not automatically the values for pure water. When flow is reported in liters per minute and heat in kilowatts, a practical form is Q (kW) = flow (L/min) × density (kg/L) × specific heat (kJ/kg·K) × ΔT (K) ÷ 60.

Here ΔT means return temperature minus supply temperature measured across the same rack-side boundary. A difference of 10°C is also a temperature interval of 10 K. The ASHRAE data-center handbook describes facility and equipment liquid cooling, including temperature classes, flow networks and pressure-drop requirements. The calculation represents heat entering the measured coolant stream, not necessarily all heat produced by the IT rack.

Worked example: a rack moving roughly 84 kW

Suppose an illustrative technology-cooling loop reports 120 L/min flow, a 25°C supply and a 35°C return. Approximate water properties as density 1.00 kg/L and specific heat 4.18 kJ/kg·K. The estimated heat is 120 × 1.00 × 4.18 × 10 ÷ 60 = 83.6 kW. The example assumes steady operation and representative, calibrated sensors.

If the same heat load passed through only 80 L/min, the simple model would require an approximately 15°C temperature rise. Higher ΔT does not automatically mean a fault; reduced flow can raise temperatures even when the total transferred heat remains similar. The safe limits depend on cold plates, coolant properties, flow distribution and server requirements. OSDCTC.016 covers what operators should capture from CDU telemetry.

Two loops and two different temperature differences

Many liquid-cooled data centers have a facility-water system and a separate technology-cooling system connected through a coolant distribution unit (CDU). Rack ΔT measures coolant warming from IT heat. CDU approach temperature, by contrast, compares the facility-side entering temperature with the technology-side coolant supply temperature. Those readings serve different purposes. A manufacturer’s heat-removal rating applies under stated flow, fluid and temperature conditions; it is not a constant promise at every operating point.

For example, Delta’s liquid-to-liquid CDU specification pairs rated kilowatts with approach temperature and loop flows. Vertiv’s CoolPhase CDU documentation shows that some architectures transfer heat into a refrigerant loop instead of a conventional facility-water loop.

Vertiv, “Hybrid Liquid Cooling Technology for Rapid AI Deployment | Vertiv CoolPhase CDU” (September 2024). Official manufacturer overview of a liquid-to-refrigerant architecture; not a field commissioning procedure.

Why real measurements depart from ideal arithmetic

Glycol mixtures change density and specific heat. Sensors with mismatched calibration, different sample times or placement near a mixing bypass can distort results. Flow measured at the CDU outlet may include branches outside the rack being assessed. Pump speed, valve position, clogged filters and air entrainment may affect distribution. Check OSDCTC.018 on quick disconnects and OSDCTC.017 on dew-point risks before recommending physical intervention.

Read-only field exercise

Record the rack identifier, coolant type, sensor locations, operating load, flow, supply and return temperatures, and timestamp. Confirm all three readings describe the same fluid circuit and sampling period. Calculate the heat estimate, compare it with a documented baseline, and inspect alerts or trend changes. Review OSDCEC.003 on cooling engineering for the larger airflow and facility context.

Never defeat leak detection, disconnect a live hose, change pump setpoints or bypass safety controls merely to improve a calculated value. Lesson conclusion: flow and ΔT can quantify heat transferred to a liquid stream, but a responsible diagnosis combines the arithmetic with actual system boundaries, manufacturer limits and safe operational evidence.

Leave a Reply