Atlas of Engineering Open technical index
Browse
HOME Home A-Z Glossary

Disciplines

18
  1. AER Aerospace
  2. ACE Automation & Control
  3. BIO Biomedical
  4. CHE Chemical
  5. CIV Civil & Construction
  6. ELC Electrical
  7. ELX Electronic
  8. PHY Physics
  9. ENV Environmental
  10. IME Industrial & Management
  11. CPE Computer
  12. MTH Mathematical
  13. MAT Materials
  14. MEC Mechanical
  15. MIN Mining & Georesources
  16. NAV Naval & Marine
  17. TEL Telecommunications
  18. ENE Energy

Exercise set

Data Center Power and Cooling Exercises

Worked engineering exercises for data center power and cooling covering IT heat load, PUE, airflow, liquid-cooling flow, UPS autonomy, redundancy, and commissioning evidence.

Branch
Energy Engineering
Content
Exercise set
Updated
Revision
v1.1.0 ยท reviewed

These exercises practise data center power and cooling calculations from an engineering point of view. The goal is not only to compute loads, flow rates, or efficiency ratios. The goal is to connect each numerical result to equipment sizing, operating margin, redundancy, commissioning evidence, and failure response.

Assume steady operation unless an exercise states otherwise. In real data centers, the final design must also consider workload ramps, power quality, harmonic current, cooling-control delay, containment leakage, water quality, maintenance bypass states, fire and leak response, sensor calibration, and the operating envelope specified by equipment manufacturers.

How to Use These Exercises

For each calculation, keep the boundary explicit:

  1. Define whether the boundary is a rack, data hall, cooling loop, UPS output, switchboard, or full facility.
  2. Separate IT load from cooling, lighting, controls, pumps, fans, and power-conversion losses.
  3. State whether the result is a normal operating value, design peak, failure case, or commissioning acceptance value.
  4. Check units before interpreting the number.
  5. Convert the numerical answer into an engineering decision.

The most common mistake is treating a data center as a single electrical load. A useful calculation must also show where the heat goes, how fast the system can respond, which component becomes limiting, and what evidence proves that the installed system behaves as expected.

For each result, state whether it supports capacity sizing, operating-envelope definition, redundancy claim, energy-performance assessment, commissioning acceptance, or a failure-response procedure. A data center margin is credible only when power, cooling, controls, and distribution evidence agree at the same boundary.

Exercise 1: IT Load and Heat Rejection

A data hall contains 24 racks. Each rack is planned for an average IT load of 18\ \text{kW} and a peak IT load of 26\ \text{kW}. The cooling design also includes 18\ \text{kW} of heat from lighting, controls, and local distribution losses inside the cooled boundary.

Find:

  1. the average IT load;
  2. the peak IT load;
  3. the peak cooling load inside the cooled boundary.

Solution

Average IT load:

P_{IT,avg}=24(18)=432\ \text{kW}

Peak IT load:

P_{IT,peak}=24(26)=624\ \text{kW}

Nearly all IT electrical power becomes heat inside the data hall. The peak cooling load inside the cooled boundary is therefore:

\dot Q_{cool,peak}=P_{IT,peak}+P_{aux,inside}
\dot Q_{cool,peak}=624+18=642\ \text{kW}

The cooling system should not be sized from the average IT load if rack deployment can reach the peak case. The average value is useful for energy estimates, but the peak value controls thermal capacity, air or liquid distribution, alarm thresholds, and failure response.

Exercise 2: Facility Power and PUE

A facility operates with an IT load of 1.20\ \text{MW}. Measured total facility power at the utility meter is 1.62\ \text{MW}.

Find the power usage effectiveness and the non-IT power demand.

Solution

Power usage effectiveness is:

\displaystyle PUE=\frac{P_{facility}}{P_{IT}}
\displaystyle PUE=\frac{1.62}{1.20}=1.35

The non-IT demand is:

P_{nonIT}=P_{facility}-P_{IT}
P_{nonIT}=1.62-1.20=0.42\ \text{MW}

The result means that 420\ \text{kW} is being used by cooling, power conversion, lighting, controls, pumps, fans, security, and other supporting systems at that operating point.

Engineering Comment

PUE is a boundary metric, not a thermal-design proof. A low PUE can still hide rack hot spots, poor redundancy, insufficient liquid-cooling flow, or weak commissioning evidence. Engineers should use PUE with rack temperature, coolant temperature, airflow, pressure, UPS efficiency, power quality, and failure-mode tests.

Exercise 3: Annual Energy Impact of a PUE Change

A data center has a steady average IT load of 850\ \text{kW}. An efficiency retrofit is expected to reduce annual average PUE from 1.48 to 1.32. Estimate the annual electrical energy reduction.

Solution

Facility power before retrofit:

P_{facility,1}=PUE_1 P_{IT}
P_{facility,1}=1.48(850)=1258\ \text{kW}

Facility power after retrofit:

P_{facility,2}=1.32(850)=1122\ \text{kW}

Power reduction:

\Delta P=1258-1122=136\ \text{kW}

Annual energy reduction:

\Delta E=\Delta P(8760)
\Delta E=136(8760)=1{,}191{,}360\ \text{kWh/year}

The retrofit saves about 1.19\ \text{GWh/year} if the average IT load and average PUE values are sustained.

Engineering Comment

This screening estimate is not enough for final business approval. A defensible retrofit case should also check seasonal operation, measurement uncertainty, chiller staging, water use, maintenance risk, control stability, redundancy during transition, and whether IT load growth changes the baseline.

Exercise 4: Airflow for a Data Hall Cooling Load

A data hall has a sensible cooling load of 520\ \text{kW}. The air-cooling design uses a supply-to-return air temperature rise of 12^\circ\text{C}. Take:

\rho_{air}=1.2\ \text{kg/m}^3
c_{p,air}=1.005\ \text{kJ/(kg K)}

Estimate the required air volume flow rate.

Solution

Use the sensible heat balance:

\dot Q=\rho \dot V c_p \Delta T

Rearrange:

\displaystyle \dot V=\frac{\dot Q}{\rho c_p \Delta T}

Use \dot Q=520\ \text{kJ/s} because 1\ \text{kW}=1\ \text{kJ/s}:

\displaystyle \dot V=\frac{520}{(1.2)(1.005)(12)}
\dot V=35.9\ \text{m}^3/\text{s}

The required airflow is approximately:

35.9\ \text{m}^3/\text{s} \times 3600 = 129{,}000\ \text{m}^3/\text{h}

Engineering Comment

This value is a theoretical heat-balance airflow. Real design must include bypass air, recirculation, containment quality, fan curves, filter loading, coil approach temperature, rack inlet limits, pressure control, and failed-fan operation. A large airflow number does not guarantee correct rack inlet temperature if the air path is poorly controlled.

Exercise 5: Liquid-Cooling Flow Rate

A direct-to-chip liquid-cooling loop removes 360\ \text{kW} from high-density racks. The coolant temperature rise across the load is limited to 8^\circ\text{C}. Use an effective coolant heat capacity:

c_p=3.8\ \text{kJ/(kg K)}

Assume coolant density is approximately:

\rho=1030\ \text{kg/m}^3

Find the required mass flow rate and volume flow rate.

Solution

Use:

\dot Q=\dot m c_p \Delta T

Rearrange:

\displaystyle \dot m=\frac{\dot Q}{c_p \Delta T}
\displaystyle \dot m=\frac{360}{(3.8)(8)}=11.84\ \text{kg/s}

Volume flow rate:

\displaystyle \dot V=\frac{\dot m}{\rho}
\displaystyle \dot V=\frac{11.84}{1030}=0.0115\ \text{m}^3/\text{s}

Convert to litres per minute:

0.0115\ \text{m}^3/\text{s} \times 60{,}000 = 690\ \text{L/min}

Engineering Comment

The flow calculation does not complete the liquid-cooling design. The engineer must also check pressure drop, pump redundancy, manifold balancing, leak detection, materials compatibility, water quality, filtration, service isolation, condensation margin, heat-exchanger approach temperature, and safe response when one pump or coolant distribution unit is unavailable.

Exercise 6: UPS Autonomy and Usable Energy

A critical IT load is 450\ \text{kW}. The design requires 10\ \text{minutes} of battery autonomy at that load. The UPS and battery path has a combined discharge efficiency of 92\%. Only 80\% of nominal battery energy is allowed to be used to protect battery life and reserve margin.

Find the minimum nominal battery energy capacity.

Solution

Required delivered energy:

E_{delivered}=P t
\displaystyle t=\frac{10}{60}=0.1667\ \text{h}
E_{delivered}=450(0.1667)=75.0\ \text{kWh}

Account for discharge efficiency:

\displaystyle E_{battery,usable}=\frac{E_{delivered}}{\eta}
\displaystyle E_{battery,usable}=\frac{75.0}{0.92}=81.5\ \text{kWh}

If only 80\% of nominal capacity is usable:

\displaystyle E_{nominal}=\frac{81.5}{0.80}=101.9\ \text{kWh}

The nominal battery capacity should be at least about:

102\ \text{kWh}

Engineering Comment

Battery autonomy calculations should be checked against end-of-life capacity, ambient temperature, discharge rate, battery management limits, maintenance bypass, generator start reliability, transfer sequence, and the actual load during the outage. A nominal kWh rating alone does not prove ride-through capability.

Exercise 7: N+1 Cooling Capacity Under Maintenance

A data hall has a design cooling load of 900\ \text{kW}. It uses four identical cooling units, each rated at 330\ \text{kW} under the expected supply and return conditions.

Check:

  1. total installed cooling capacity;
  2. available capacity with one unit out of service;
  3. whether the system meets the 900\ \text{kW} load with one unit unavailable.

Solution

Total installed capacity:

Q_{installed}=4(330)=1320\ \text{kW}

Capacity with one unit unavailable:

Q_{N+1}=3(330)=990\ \text{kW}

Margin during one-unit outage:

Q_{margin}=990-900=90\ \text{kW}

Percentage margin:

\displaystyle \frac{90}{900}\times 100=10\%

The system meets the nominal N+1 capacity check with a 10\% margin.

Engineering Comment

This is only a capacity-screening result. The real N+1 claim must also prove that the remaining units can deliver air or coolant to the right racks, controls remain stable, electrical feeds can support the failed-unit state, valves and dampers move correctly, and no single distribution failure removes more capacity than assumed.

Exercise 8: Commissioning Evidence for a Cooling Step Test

During commissioning, a data hall step test increases IT load from 300\ \text{kW} to 520\ \text{kW}. The measured increase in facility power is 305\ \text{kW} after cooling controls settle.

Find the incremental facility power per incremental IT power and interpret the result.

Solution

Incremental IT load:

\Delta P_{IT}=520-300=220\ \text{kW}

Incremental facility power:

\Delta P_{facility}=305\ \text{kW}

Incremental ratio:

\displaystyle R=\frac{\Delta P_{facility}}{\Delta P_{IT}}
\displaystyle R=\frac{305}{220}=1.39

For this operating step, each additional kilowatt of IT load increased facility power by about 1.39\ \text{kW}.

Engineering Comment

This ratio is not the same as annual PUE, but it is useful commissioning evidence. It shows how the facility responds to an incremental load at a specific operating point. Engineers should compare the result with expected fan, pump, chiller, economizer, UPS, and control behaviour. If the ratio is much higher than expected, likely causes include poor containment, unnecessary chiller staging, unstable control loops, high fan power, or measurement boundary error.

Review Checklist

Before accepting a data center power and cooling calculation, check:

  • whether the boundary is clearly defined;
  • whether load values are average, peak, diversified, or failure-case values;
  • whether all units are consistent;
  • whether electrical power has been translated into heat at the correct boundary;
  • whether airflow or coolant flow reaches the rack, not only the plant;
  • whether redundancy is proven for distribution as well as capacity;
  • whether PUE is supported by temperature, flow, pressure, power-quality, and reliability evidence;
  • whether maintenance bypasses, single failures, workload ramps, sensor calibration, and control-loop delays are included in acceptance evidence;
  • whether commissioning step tests verify rack inlet conditions and coolant or airflow delivery, not only plant capacity;
  • whether commissioning tests measure the same quantities assumed in design.

Good data center engineering turns power and cooling calculations into operating proof. A correct arithmetic result is only the first step; the design is credible when the installed facility can demonstrate the same behaviour under load, maintenance, fault, and recovery conditions.

REF

See also

  1. Data Center Power and Cooling Engineering Energy Engineering
  2. Data Center Power and Cooling Formula Sheet Energy Engineering
  3. Thermal Management and Heat Transfer Exercises Mechanical Engineering
  4. Data Center Cooling Load Estimation Project Energy Engineering
  5. How Liquid Cooling Works in Data Centers Energy Engineering
  6. AI Data Center Grid Connection Case Study Energy Engineering
  7. Building Energy Systems and HVAC Performance Energy Engineering
  8. Thermal Management, Heat Transfer, and Cooling Systems Mechanical Engineering
  9. Thermal Energy Systems and Heat Exchangers Energy Engineering
  10. Energy Efficiency and Heat Recovery Systems Energy Engineering
  11. Electrical Power Distribution, Substations, and Protection Coordination Electrical Engineering
  12. AC Power Systems Electrical Engineering
  13. Power Electronics and Converter Systems Electrical Engineering
  14. Power Usage Effectiveness Energy Engineering metric
  15. Rack Power Density Energy Engineering metric
  16. Liquid Cooling Energy Engineering method
  17. Uninterruptible Power Supply Electrical Engineering device
  18. Power Electrical Engineering quantity
  19. Efficiency Energy Engineering metric
  20. Heat Flux Energy Engineering quantity
  21. Junction Temperature Electronic Engineering quantity
  22. Thermal Stress Mechanical Engineering phenomenon
  23. Heat Exchanger Energy Engineering device
  24. Flow Rate Mechanical Engineering quantity
  25. Density Materials Engineering quantity
  26. Viscosity Mechanical Engineering quantity
  27. Reynolds Number Mechanical Engineering quantity
  28. Power Factor Electrical Engineering metric
  29. Harmonic Distortion Electrical Engineering metric
  30. Overcurrent Protection Electrical Engineering device
  31. Closed-Loop Control Automation and Control Engineering concept
  32. Reliability Industrial and Management Engineering metric
  33. Failure Mode Industrial and Management Engineering concept
  34. Validation Industrial and Management Engineering method
  35. Uncertainty Analysis Mathematical Engineering method
Branches