Dissertation/Thesis Abstract

Repairable file and storage systems
by Zhu, Ningning, Ph.D., State University of New York at Stony Brook, 2008, 239; 3358208
Abstract (Summary)

The data contents of an information system may be corrupted due to human errors, malicious attacks or untrusted software. The financial loss of such corruption is typically proportional to the amount of time required to recover the system's data/service. Recognizing that it is impossible to build absolutely secure computer systems and that human errors are inevitable, this dissertation proposes a repairable system framework which greatly reduces data loss and system downtime with minimum cost and performance penalty. We illustrate that repairability is affordable and ready to be integrated into main stream file/storage system.

Repairable file/storage system needs to perform two tasks. First it has to maintain all the raw data so that every update is undoable. Secondly it has to keep track of data updates due to errors and attacks so that only the data affected by mistakes or attacks are rolled back to their last known consistent state. We develop two novel comprehensive versioning schemes for repairable NFS file server and for repairable SAN storage system. We design a simple solution for dependency tracking and integrate it with both schemes. We also developed an NFS trace play toolkit and gained experience on trace driven file system evaluation.

For the repairable file system, we focus on the performance optimization in the absence of failures and errors. Empirical measurements show that the performance overhead due to repairability is less than 10%. For the repairable storage system, more focuses are shifted to the integration with traditional fault tolerance techniques. Evaluation results show that the repairable storage system is available upon any single point of failure, including disk failure with and without data losses, power failure, network failure and software crash failure.

Indexing (document details)
Advisor: Chiueh, Tzi-cker, Zadok, Erez
School: State University of New York at Stony Brook
School Location: United States -- New York
Source: DAI-B 70/05, Dissertation Abstracts International
Subjects: Computer science
Keywords: Data recovery, Dependency tracking, Fault tolerance, File systems, Repairability, Storage systems, Versioning
Publication Number: 3358208
ISBN: 9781109168730
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy