Dissertation/Thesis Abstract

A Novel Integrated Fuzzy Control System Towards Automated Local Airflow Management in Data Centers
by Mohsenian, Ghazal, M.S.M.E., State University of New York at Binghamton, 2020, 72; 28029188
Abstract (Summary)

Data Centers (DCs) have become an indispensable part of modern computing infrastructures. Today, DCs are dynamic environments, with considerable fluctuations in workload and power dissipation. As a result, active monitoring and holistic management of all infrastructure assets are essential. It is vital to ensure Information Technology Equipment (ITE) has access to sufficient air (provisioned) at a proper temperature for their optimal and continuous operation. Hot air recirculation, elevated fan speed, and hot spots are among known consequences of an under-provisioned cold aisle. On the other hand, over-provisioning a cold aisle can lead to a significant energy loss due to cooling air bypass. In most air-cooled data centers, the required airflow for cooling of the ITE is delivered from cooling units to the raised floor plenum which flows to server racks through perforated floor tiles in the cold aisles. Generally, the perforated tiles have fixed openings and are not adaptive to the airflow demand and ITE load changes in the server racks of each aisle. The number of active servers in an aisle and their workload levels may be varied by load balancers in a DC depending on the processing demands of the IT hardware at a given time. Simultaneous manual tuning of the airflow at the floor tile level is impossible or at least impractical. To manage cold air delivery to individual aisles based on airflow demand, variable airflow panels (air dampers) are deployed. In this arrangement, each air damper is controlled to go from fully opened to fully closed.

The presented work in this thesis introduces an automated dynamic airflow management technique using air dampers. First, the experimental data is gathered to characterize airflow and pressure across the air dampers. Then, an Artificial Neural Network (ANN) based on the experimental data is developed for careful and precise damper characterization. Afterward, a fuzzy controller is designed and optimized to regulate the local airflow rate to the aisles based on the Differential Pressure (ΔP) between the Cold Aisle Containment (CAC) and the room. The control system regulates airflow delivery during workload changes in a DC by adjusting the air dampers’ Open Area Ratio (OAR). This study experimentally examines several scenarios for improving the thermal management of DCs with automatic controls.

Indexing (document details)
Advisor: Sammakia, Bahgat G.
Commitee: Murray, Bruce T., Chiarot, Paul R.
School: State University of New York at Binghamton
Department: Mechanical Engineering
School Location: United States -- New York
Source: MAI 82/4(E), Masters Abstracts International
Source Type: DISSERTATION
Subjects: Energy, Mechanical engineering, Computer Engineering
Keywords: Active control, Air damper, Airflow management in data centers, Artificial neural network, Automated data center, Dynamic workload in data centers
Publication Number: 28029188
ISBN: 9798678160256
Copyright © 2021 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest