Graduation Semester and Year

2016

Language

English

Document Type

Thesis

Degree Name

Master of Engineering in Mechanical Engineering

Department

Mechanical and Aerospace Engineering

First Advisor

Dereje Agonafer

Abstract

The information technology (IT) owners are experiencing a greater cooling challenge because of the increase in power density due to modern computational needs. The non-uniform power density in each server is forcing the industry to use hybrid cooling technology. Server components of different cooling requirement needs air water hybrid cooling which offers variable design alternatives. Such hybrid cooling technology cools the high heat generating components by using water or water based fluid, whereas, the rest of the components are cooled by air using internal fans. Conventional air cooling is more than sufficient for the components with less thermal demand. Air cooling is cheap, highly available and it has better serviceability than any other cooling methods. The objective is to optimize the cooling power of the air cooling loop of such hybrid cooled server. As the major components are cooled by the water based fluid, the other components generate less heat which can be cooled by much less volume of air then supplied in air cooled server. The volume of air supplied is controlled by varying the air flow rate through the internal fans. Also number of fan was reduced to 3 instead of 5 to minimize the power consumption. Parameters like CPU and memory utilization are varied with the flow rate. ASHRAE recommends that the most data centers can be maintained between 20 and 25°C, with an allowable range of 15 to 32°C. But for this type of hybrid cooling servers, the processor is cooled by water. So the servers can operate at much higher inlet air temperature. In this paper the hybrid cooled servers will be characterized also. The server used for experimental testing has processor with 135 watt thermal design power. Also, the server utilizes distributed pumping i.e. each cold plate has its own pump. The test matrixes consider supply and return water temperatures, flow rate of coolant for optimizing the cooling power consumption. The supply inlet water temperature was varied by LabView code. Further, processor and outlet temperature was monitored for better understanding the case scenario. The relation between supply water temperature and different power utilization gives the data for modeling different cooling infrastructure. This in turn, will give an idea of power savings by utilizing such energy efficient hybrid solution for cooling servers in a datacenter.

Keywords

Data center, Thermal management

Disciplines

Aerospace Engineering | Engineering | Mechanical Engineering

Comments

Degree granted by The University of Texas at Arlington

Download

Share

COinS