A Comparative CFD Study of Two Air Distribution Systems with Hot Aisle Containment in High-Density Data Centers

Overview

The study asks how the thermal performance of raised-floor and hard-floor air distribution differs with HAC in high-density data centres, and how that difference changes when a cooling unit fails. The IT room has 40 A-type racks at 16 kW and 60 B-type racks at 8.8 kW, 1830 mm cold aisles and 1220 mm hot aisles, ducted HAC, six CRAHs (234,000 m³/h total) supplying 15 °C and a 15 °C server delta-T. Both layouts were tested with all six units running and with the unit serving the high-density rows failed (N-1).

Under normal cooling, mean inlet temperatures were 16.5 °C (raised floor) and 16.9 °C (hard floor), but maximum inlet temperatures were 20.6 °C and 30.2 °C. Only 97% of racks met the ASHRAE recommended range on the hard floor, with failures in the racks nearest the CRAHs. Maximum rack recirculation ratio was 25.7% on the raised floor versus 50% on the hard floor, meaning some hard-floor racks received only half their required airflow. On the hard floor, supply velocities of 10 m/s or more persisted to the eighth rack position, versus the fourth with a raised floor, which explains the starvation of near-CRAH racks.

With one CRAH failed (N-1), the remaining five kept almost normal conditions on the raised floor. On the hard floor, more racks had recirculation above 40-50%, with inlet temperatures of about 26.1-30.4 °C. The authors report that raised-floor HAC improved air distribution efficiency by 28% over hard-floor HAC, equivalent to a 40% reduction in racks with recirculation above 20%.

Limits: the model is not validated against field measurements, and results depend on one reference hall geometry, CRAH layout and 15 °C supply. The conclusion is that HAC systems with a raised floor have higher cooling efficiency than those with a hard floor.

Key points

1. IT room: 40 A-type racks at 16 kW and 60 B-type racks at 8.8 kW, 1830 mm cold aisles, 1220 mm hot aisles, ducted HAC, six CRAHs (234,000 m³/h total) supplying 15 °C, 15 °C server delta-T.
2. Under normal cooling, mean inlet temperatures were 16.5 °C (raised floor) and 16.9 °C (hard floor); maximum inlet temperatures were 20.6 °C and 30.2 °C; only 97% of racks met the ASHRAE recommended range on the hard floor.
3. Maximum rack recirculation ratio was 25.7% on the raised floor versus 50% on the hard floor; some hard-floor racks received only half their required airflow.
4. On the hard floor, supply velocities of 10 m/s or more persisted to the eighth rack position, versus the fourth with a raised floor.
5. With one CRAH failed (N-1), the remaining five kept almost normal conditions on the raised floor; on the hard floor more racks had recirculation above 40-50%, with inlet temperatures of about 26.1-30.4 °C.
6. The authors report raised-floor HAC improved air distribution efficiency by 28% over hard-floor HAC, equivalent to a 40% reduction in racks with recirculation above 20%.

Using it in design

• In high-density HAC halls, prefer a raised floor to hard-floor supply; on a hard floor high-velocity jets can starve the racks nearest the CRAHs and draw hot air back into them.
• Run cooling unit failure (N-1) as a separate scenario in thermal verification and report recirculation ratio and maximum inlet temperature by rack.
• For hard-floor supply, add CRAH discharge velocity and rack distance (10 m/s jets persisting to the eighth rack) to the design check.

Who should read it

Cooling designers choosing between raised-floor and hard-floor air distribution for high-density data halls.

Document contents

1. 1. Introduction
2. 2. Methodology (2.1 Analytic Validity of CFD Simulation for Data Centers, 2.3 Basic Analysis Model: IT server room, server, hot aisle containment; 2.4 CFD Analysis Alternatives)
3. 3. Numerical Analysis Results (3.1 Normal Cooling Conditions (ALT-1 and ALT-2): temperature distribution, airflow distribution, inlet temperature, recirculation ratio, operational net cooling power; 3.2 Cooling fault conditions: same measures)
4. 5. Conclusions

Related standards

• ASHRAE TC 9.9 Thermal Guidelines for Data Processing Environments

Read the document

Publication details

Authors

Jinkyun Cho; Jesang Woo; Beungyong Park; Taesub Lim

Publisher

Energies (MDPI AG)

Year

2020

Volume

13

Issue

22

Pages

6147

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/en13226147

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation