Optimization of Airflow Organization in Bidirectional Air Supply Data Centers in China

Overview

The study asks how to improve airflow organisation in a bidirectional air-supply data centre in a hot-summer, warm-winter region. The IT room is 502.3 m² with a 0.7 m raised floor and 4.5 m clear height; nine CRACs (seven running, two standby), each rated 34,800 m³/h, supply under the floor with side return to 142 racks in eight rows on 1.2 m aisles. Validation used 72 temperature points (0.5, 1.2 and 1.8 m heights) and 144 tile-velocity points: inlet temperature MBE was 0.70 °C with RMSE 0.92 °C, outlet MBE 0.67 °C, RMSE 1.46 °C, and tile airflow RMSE 0.011 m³/s.

Model settings: standard k-epsilon, polyhedral mesh of 6,972,368 cells, perforated tile porosity 36% and rack door porosity 86% as porous jumps, and server heat as volumetric sources from measured PDU power. Colliding plenum jets gave peak static pressures of 6-14 Pa at the collision interfaces. Mean tile flow was about 0.20 m³/s in the one-sided supply zone and 0.25 m³/s in the two-sided zone, with minimums of 0.14 m³/s in rows AE and AF. The cooling system over-supplies air and cold-air bypass dominates, yet severe mixing of cold and hot air gives an imbalanced cooling capacity distribution.

Seven retrofit cases were scored with RTI, SHI and RCI-HI: three containment variants (doors only, roof only, full), blanking panels, and blanking combined with each containment variant. Baseline RTI was 0.786, SHI 0.230 and RCI-HI 0.992. Blanking panels alone gave SHI 0.132 and RCI-HI 1; blanking plus full cold-aisle containment (Case 7) gave RTI 1 and SHI 0.027. SHI reductions versus baseline were: door-only containment 1.30%, roof-only 0.43%, full containment alone 12.17%, blanking alone 42.61% and blanking plus full CAC 88.26%.

Blanking was the best single measure; blanking plus full containment was best overall, and blanking plus roof-only containment was the cheaper runner-up. Containment without blanking achieved little. Limits: a single hall and climate zone; results depend on this hall's geometry and CRAC layout.

Key points

1. The IT room is 502.3 m² with a 0.7 m raised floor and 4.5 m clear height; nine CRACs (seven running, two standby), each 34,800 m³/h, serve 142 racks in eight rows on 1.2 m aisles.
2. Validation used 72 temperature points and 144 tile-velocity points: inlet temperature MBE 0.70 °C, RMSE 0.92 °C; outlet MBE 0.67 °C, RMSE 1.46 °C; tile airflow RMSE 0.011 m³/s.
3. Model: standard k-epsilon, polyhedral mesh of 6,972,368 cells, tile porosity 36% and rack door porosity 86% as porous jumps, server heat as volumetric sources from measured PDU power.
4. Colliding plenum jets gave peak static pressures of 6-14 Pa; mean tile flow about 0.20 m³/s in the one-sided zone and 0.25 m³/s in the two-sided zone, minimum 0.14 m³/s in rows AE and AF.
5. Baseline RTI 0.786, SHI 0.230, RCI-HI 0.992. Blanking alone: SHI 0.132, RCI-HI 1. Blanking plus full CAC (Case 7): RTI 1, SHI 0.027.
6. SHI reductions versus baseline: door-only containment 1.30%, roof-only 0.43%, full containment alone 12.17%, blanking alone 42.61%, blanking plus full CAC 88.26%; containment without blanking achieved little.

Using it in design

• In retrofitting an existing raised-floor hall, first close unused rack spaces with blanking panels (alone they cut SHI by 42.61%); containment alone gave little benefit.
• In bidirectional plenum supply, check by CFD or measurement the static pressure imbalance (6-14 Pa) and uneven tile flow (0.14 to 0.25 m³/s) where opposing jets collide.
• If budget is limited, compare blanking plus roof-only containment as the cheaper option and blanking plus full cold-aisle containment as the best option, scoring both with RTI, SHI and RCI-HI.

Who should read it

Cooling engineers and operators improving airflow in existing raised-floor data halls.

Document contents

1. 1. Introduction
2. 2. Materials and Methods (2.1 Field Experiment: overview of data centers, experimental scheme; 2.2 Numerical Simulation: governing equations, simplification assumptions, boundary conditions; 2.3 Thermal Metric)
3. 3. Results (3.1 Model Validation, 3.2 Air Supply Characteristics, 3.3 Flow Field Distribution, 3.4 Temperature)
4. 4. Airflow Organization Optimization (4.1 Airflow Optimization Strategies, 4.3 Analysis)
5. 5. Conclusions

Related standards

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

Read the document

Publication details

Authors

Yixin Wu; Junwei Yan; Xuan Zhou

Publisher

Applied Sciences (MDPI AG)

Year

2025

Volume

15

Issue

10

Pages

5711

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/app15105711

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation