Date of Award

5-2012

Degree Type

Dissertation

Embargo Date

8-17-2012

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical and Aerospace Engineering

Advisor(s)

H. Ezzat Khalifa

Keywords

Data Center, Energy Efficiency, Virtualization

Subject Categories

Mechanical Engineering

Abstract

This work investigates the practical implementation of so-called thermally aware, energy optimized load placement in air-cooled, raised floor data centers to reduce the overall energy consumption, while maintaining the reliability of the IT equipment. The work takes a systematic approach to modeling the data center's airflow, thermodynamic and heat transfer characteristics - beginning with simplified, physics-inspired models and eventually developing a high-fidelity, experimentally validated thermo-hydraulic model of the data center's cooling and power infrastructure. The simplified analysis was able to highlight the importance of considering the trade-off between low air supply temperature and increased airflow rate, as well as the deleterious effect of temperature non-uniformity at the inlet of the racks on the data center's cooling infrastructure power consumption. The analysis enabled the development of a novel approach to reducing the energy consumption in enclosed aisle data centers using bypass recirculation. The development and experimental validation of a high-fidelity thermo-hydraulic model proceeded using the insights gained from the simple analysis. Using these tools, the study of optimum load placement is undertaken using computational fluid dynamics as the primary tool for analyzing the complex airflow and temperature patterns in the data center and is used to develop a rich dataset for the development of a reduced order model using proper orthogonal decomposition. The outcome of this work is the development of a robust set of rules that facilitate the energy efficient placement of the IT load amongst the operating servers in the data center and operation of the cooling infrastructure. The approach uses real-time temperature measurements at the inlet of the racks to remove IT load from the servers with the warmest inlet temperature (or add load to the servers with the coldest inlet temperature). These strategies are compared to conventional load placement techniques and show superior performance by considering the holistic optimization of the data center and cooling infrastructure for a range of data center IT utilization levels, operating strategies and ambient conditions.

Access

Open Access