Dissertation/Thesis Abstract

Using content addressable techniques to optimize client -server systems
by Tolia, Niraj, Ph.D., Carnegie Mellon University, 2007, 173; 3286685
Abstract (Summary)

Efficient access to bulk data over the Internet has become a critical problem in today's world. Even while bandwidth, both in the core and the edge of the network, is improving, the simultaneous growth in the use of digital media and large personal data sets is placing increasing demands on it. Further, with increasing trends towards mobility, an increasing amount of data access is over cellular and wireless networks. These trends are placing pressure on applications and the Wide-Area Network (WAN) to deliver better performance to the end user.

This dissertation puts forward the claim that external resources can be used in an opportunistic manner to optimize bulk data transfer in WAN-based client-server systems. In particular, it advocates the use of content-addressable techniques, a system-independent method for naming objects, to cleanly enable these optimizations. By detecting similarity between different data sources, these techniques allow external sources to be used without weakening any attributes, including consistency, of legacy systems and with no or minor changes to the original system.

This dissertation validates this claim empirically through the use of five case studies that encompass the two traditional forms of data storage, file systems and database systems, and then generalizes the claim in the form of a generic transfer service that can be shared by different applications. In each of these case studies, I focus on three questions. First, even with the addition of content-based optimizations, how does the resulting system still maintain the attributes and semantics of the original system? Second, how do these systems efficiently find similarity in data? Third, does the use of these content-addressable techniques provide a substantial performance improvement when compared to the original system? These questions are answered with a detailed description of the system design and implementation and a careful evaluation of the prototypes with both synthetic and real benchmarks.

Indexing (document details)
Advisor: Satyanarayanan, M.
School: Carnegie Mellon University
School Location: United States -- Pennsylvania
Source: DAI-B 68/10, Dissertation Abstracts International
Subjects: Computer science
Keywords: Client-server, Content-addressable storage, Database systems, Distributed file systems, Mobility, Network optimizations, Portable storage
Publication Number: 3286685
ISBN: 978-0-549-29868-7
Copyright © 2021 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy