Evaluation and Optimization of a Two-Phase Liquid-Immersion Cooling System for Data Centers

Overview

The study starts from the premise that an efficient cooling system for data centres can boost server working efficiency and promote energy savings, and explores the performance of a two-phase cooling system with a laboratory experiment and CFD simulation. The working fluid is Novec 7100. The COP and partial PUE (pPUE) were evaluated under various IT loads; the relationship between the interval of submerged servers and their surface temperatures was then evaluated by CFD to obtain the minimum intervals keeping server surfaces below 85 °C.

As server power rose from 1127 W to 1577 W, system COP increased from 19.0 to 26.7, generally 4-20 units higher than reported air- or water-cooled systems. The pPUE of the two-phase system fell from 1.053 to 1.037 as server power increased. Exergy efficiency ranged from 12.65% to 18.96% (average 14.84%), with most exergy destruction on the tank side. COP and PUE depended on room temperature and on the fan and pump speed of the external heat exchanger.

To keep server surfaces below 85 °C, server spacing must be at least 15 mm at 1000 W and 20 mm at 1500 W; at 2000 W or 3000 W even 30 mm was insufficient. For a given tank, a closer interval between servers causes higher server surface temperature, and the allowable board pitch grows quickly with per-server power.

Limits: a single laboratory system, one fluid (Novec 7100) and a 1.1-1.6 kW experimental range; spacings above 2000 W were predicted by CFD. Results depend on tank geometry and the external heat-exchanger setting.

Key points

1. As server power rose from 1127 W to 1577 W, system COP increased from 19.0 to 26.7, generally 4-20 units higher than reported air- or water-cooled systems.
2. The pPUE of the two-phase system fell from 1.053 to 1.037 as server power increased.
3. Exergy efficiency ranged from 12.65% to 18.96% (average 14.84%), with most exergy destruction on the tank side.
4. To keep server surfaces below 85 °C, server spacing must be at least 15 mm at 1000 W and 20 mm at 1500 W; at 2000 W or 3000 W even 30 mm was insufficient.
5. COP and PUE depended on room temperature and on the fan and pump speed of the external heat exchanger.
6. The working fluid was Novec 7100.

Using it in design

• In a two-phase immersion tank, set server spacing by power: at least 15 mm at 1000 W and 20 mm at 1500 W for the 85 °C surface limit, and wider for 2000 W and above.
• Report efficiency (COP 19-26.7, pPUE 1.037-1.053) by IT load; efficiency improved as load rose, so do not expect design-point values at low load.
• Manage the external heat exchanger's fan and pump speed and room temperature as control variables for COP and PUE.

Who should read it

Cooling engineers designing two-phase immersion tanks and server spacing.

Document contents

1. 1. Introduction
2. 2. Methodology (2.1 Experimental Study, 2.2 CFD Simulation: physical model and boundary conditions, CFD modeling of boiling flow, material properties, grid independence, CFD model validation; 2.3 COP and pPUE, 2.4 Exergy Analysis)
3. 3. Results and Discussion (3.1 COP and pPUE, 3.2 Exergy Analysis, 3.3 Simulation Results)
4. 4. Conclusions

Read the document

Publication details

Authors

Cheng Liu; Hang Yu

Publisher

Energies (MDPI AG)

Year

2021

Volume

14

Issue

5

Pages

1395

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/en14051395

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation