Abstract

Data Centers have become the digital backbone of the modern society with the advent of cloud computing, social networking, big data analytics etc. They play a vital role in processing a large amount of information generated. The number of data centers and the servers present in them have been on the rise over the last decade. This has eventually led to the increase in the power consumption of the data center due to the power-hungry interconnect fabric which consists of switches, routers and switching fabric necessary for communication in the data center. Moreover, a major portion of the power consumed in a data center belongs to cooling infrastructure. The data center’s complex cabling prevents the heat dissipation by obstructing the air flow resulting in the need for a cooling infrastructure. Additionally, the complex cabling in traditional data centers poses design and maintenance challenges. In this work, these problems of traditional data centers are addressed by designing a unique new server-to-server wireless Data Center Network (DCN) architecture.

The proposed design methodology uses 60GHz unlicensed millimeter-wave bands to establish direct communication links between servers in a DCN without the need for a conventional fabric. This will reduce the power consumption of the DCN significantly by getting rid of the power-hungry switches along with an increase in the independency in communication between servers.

In this work, the previous traffic models of a data center network are studied and a new traffic model very similar to the actual traffic in a data center is modeled and used for simulating the DCN environment. It is estimated that the proposed DCN architecture’s power consumption is lowered by six to ten times in comparison to the existing conventional DCN architecture. Having established the power model of a server-to-server wireless DCN in terms of its power consumption, we demonstrate that such a power-efficient wireless DCN can sustain the traffic requirements encountered and provide data rates that are comparable to traditional DCNs. We have also compared the efficiency and performance of the proposed DCN architecture with some of the other novel DCN architectures like DCell, BCube with the same traffic.

Library of Congress Subject Headings

Wireless sensor networks--Energy conservation; Computer network architectures; Data libraries; Data warehousing; Web servers

Publication Date

11-2017

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Engineering (MS)

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Amlan Ganguly

Advisor/Committee Member

Andres Kwasinski

Advisor/Committee Member

Mineosk Kwon

Campus

RIT – Main Campus

Plan Codes

CMPE-MS

Share

COinS