Airflow management and energy saving potentials at a high-density data center with stepped-like server placement

ABSTRACT

This paper proposed and introduced an innovative concept of step-like server placement in data centers (DCs), replacing the traditional servers placed on the same vertical plane with step-like server placement. The studied simplified DC model was established based on an actual operating DC. The feasibility and reliability of the simplified model was validated by on-site experimental results. This paper is analyzed two scenarios: the effects of horizontal spacings (0, 0.01, 0.02, 0.03, and 0.04 m) between adjacent servers and with the effect of different supply air temperatures (SATs) (22.5, 23, 23.5, and 24°C) on the rack hotspot. The results show that step-like server placement can effectively enhance the thermal distribution, while the optimum thermal distribution is achieved with the spacing of 0.01 m in case 2. Under this circumstance, the rack hotspot reduced by 2.5°C, while case 2 is selected for further analysis in scenario 2. In scenario 2, the thermal environment in cases 6–8 are totally better than that in case 1 in terms of rack hotspot and temperature distribution. Although there appears another moderate heat accumulation at the top-side, the overall thermal distribution in the case with SAT of 24°C is still better than that in original case. The electricity consumptions of the CRACs can be saved by approximately 146 and 195 kWhr/day with the most optimal configuration.

KEYWORDS:

• Data center
• energy saving
• rack-level airflow management
• step-like server placement
• temperature distribution
• thermal environment

Abbreviations

CAC Cold aisle containment

COP Coefficient of performance

CPU Central processing unit

CRACs Computer room air-conditioning units

DC Data center

DES Detached Eddy Simulations

HVAC Heating, ventilation, and air conditioning

MTP Measuring temperature point

MVP Measuring velocity point

OHA Open hot aisle

RSM Reynolds Stress Model

SAT Supply air temperature

STP Simulated temperature point

SVP Simulated velocity point

TP Temperature point

UFAD Under-floor air distribution

Nomenclature

W₁ Length of each server, m

C₁ Distance between the rack rear door and the terminal of the bottom server, m

C₂ Distance between the rack front door and front vertical side of the top server, m

D Distance between the front vertical sides of the bottom and the top servers, m

L₁ Length of each rack.

$Q_1$ Daily cooling energy saving, kWhr

$Q_2$ Daily total electricity saving for CRAC units, kWhr

$cp$ Specific heat capacity of air, kJ/(kg K)

$m$ Mass of air, kg

$\rho$ Air density, kg/m³

$V$ Volume of air from the CRAC units, m³

$v$ Air velocity through the air outlet of the CRAC units, m/s

$A$ Area of air vent of each CRAC, m²

$\mathrm{\Delta }t$ SAT difference between the original and optimal model, K

$SA{T}_{C1}$ SAT of case 1, °C

$SA{T}_{C9}$ SAT of case 9, °C

$SA{T}_{Cx}$ SAT of the selected case, °C

x Number of the optimum case

$CO{P}_{CRAC}$ COP of CRAC units

$\vec{u}$ Average velocity vector

$p$ Static pressure, kPa

$T$ Static temperature, kPa

$\vec{g}$ Gravitational acceleration vector

$v_{eff}$ Effective fluid viscosity

$k_{eff}$ Thermal conductivity

$S$ Volumetric heat sources

U Unit of server dimension

Acknowledgments

We also want to acknowledge Information Center in Jiangpu Campus of Nanjing Tech University, Nanjing, China, for providing the experimental site.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [51978481]; 13th Five-year” National Key R&D Program of China [2017YFC0702200].