Optimization Strategies for Underfloor Air Distribution in a Small-Scale Data Center

Overview

The study asks how underfloor air distribution in a small data centre can be optimised, given that hot and cold airflow mixing in rapidly expanding 5G-era data centres keeps the thermal environment from meeting operating requirements. The room is 14.75 m × 13.75 m × 3.45 m with three units of 6.5 m³/s each (60 kW total, 33.4 kW sensible), five cabinet rows (13-14 cabinets each), a 0.45 m raised floor and 0.6 m × 0.6 m tiles with 45% open area; design supply is 18 °C. Validation errors: tile outlet temperature about 2% on average, tile velocity 18% (12% excluding outliers), and overall averages of 9.44% for temperature and 12.6% for velocity.

At 300, 400 and 500 mm plenum heights, minimum grille flows were negative: -0.044, -0.02 and -0.024 m³/s respectively. At 700 mm the grille flow range was 0.081-0.354 m³/s and at 800 mm 0.155-0.34 m³/s. The best plenum height range is 600-800 mm, with little improvement above 800 mm. At 600 mm, maximum cabinet inlet and outlet temperatures were 19.3 °C and 34 °C and RHI was 0.9622.

RHI was 0.9622 open, 0.9710 with cold aisles closed, 0.9271 with hot aisles closed and 0.9975 with both closed; hot-aisle-only containment let supply air bypass to the unit returns, and the highest temperature reached 40.3 °C. Cold-aisle containment allowed the supply air temperature to rise from 18 °C to 20 °C, with average cabinet inlet and outlet temperatures of 18-18.4 °C and 29-30 °C.

Limits: a single small room, with a 12.6% average velocity validation error; results depend on this room's geometry and unit layout. Hot-aisle containment lowered RHI in this room, so the choice of containment type should be verified room by room.

Key points

1. Room 14.75 m × 13.75 m × 3.45 m with three units of 6.5 m³/s each (60 kW total, 33.4 kW sensible), five cabinet rows (13-14 cabinets each), a 0.45 m raised floor and 0.6 m × 0.6 m tiles with 45% open area; design supply 18 °C.
2. Validation errors: tile outlet temperature about 2% on average, tile velocity 18% (12% excluding outliers), overall averages of 9.44% for temperature and 12.6% for velocity.
3. At 300, 400 and 500 mm plenum heights, minimum grille flows were negative: -0.044, -0.02 and -0.024 m³/s; at 700 mm the range was 0.081-0.354 m³/s and at 800 mm 0.155-0.34 m³/s.
4. The best plenum height range is 600-800 mm, with little improvement above 800 mm; at 600 mm, maximum cabinet inlet and outlet temperatures were 19.3 °C and 34 °C and RHI was 0.9622.
5. RHI was 0.9622 open, 0.9710 with cold aisles closed, 0.9271 with hot aisles closed and 0.9975 with both closed; hot-aisle-only containment let supply air bypass to the unit returns.
6. Cold-aisle containment allowed the supply air temperature to rise from 18 °C to 20 °C.

Using it in design

• Choose a raised-floor height of at least 600 mm; at 300-500 mm some grilles had reverse (negative) flow and uneven static pressure, and gains levelled off above 800 mm.
• Prefer cold-aisle containment and consider raising supply from 18 to 20 °C; in this room hot-aisle-only containment lowered RHI because cold air bypassed to the unit returns.
• Verify plenum height and containment decisions room by room with tile airflow measurement and CFD.

Who should read it

Cooling engineers evaluating underfloor air distribution and aisle containment options in small server rooms.

Document contents

1. 1. Introduction
2. 2. Research Methodology (2.1 Experimental Testing: establishment of geometric model, description of experimental testing; 2.2 Numerical Method and Boundary Conditions; 2.3 Thermal Environment Evaluation Indicators)
3. 3. Results and Discussion (3.1 Validation of Experimental and Simulation Results, 3.2 Analysis of Data Center Thermal Environment Simulation Results)
4. 4. Airflow Distribution Optimization Strategy (4.1 Raised Floor Height Adjustment: airflow volume, pressure field, temperature field, evaluation indicators; 4.2 Closed Cold/Hot Aisles: streamline analysis, evaluation indicators, impact of increased air supply temperature)
5. 5. Conclusions

Related standards

• GB 50174 Code for design of data centers
• ASHRAE TC 9.9 Thermal Guidelines for Data Processing Environments

Read the document

Publication details

Authors

Fengjiao Yu; Hongbing Chen; Wenqian Wang; Jingjing An

Publisher

Buildings (MDPI AG)

Year

2025

Volume

15

Issue

3

Pages

428

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/buildings15030428

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation