Visible to the public File Defragmentation Scheme for a Log-Structured File System

TitleFile Defragmentation Scheme for a Log-Structured File System
Publication TypeConference Paper
Year of Publication2016
AuthorsPark, Jonggyu, Kang, Dong Hyun, Eom, Young Ik
Conference NameProceedings of the 7th ACM SIGOPS Asia-Pacific Workshop on Systems
Date PublishedAugust 2016
PublisherACM
Conference LocationNew York, NY, USA
ISBN Number978-1-4503-4265-0
Keywordscleaning, file defragmentation, Log-structured file systems, pubcrawl170201
Abstract

In recent years, many researchers have focused on log-structured file systems (LFS), because it gracefully enhances the random write performance and efficiently resolves the consistency issue. However, the write policy of LFS can cause a file fragmentation problem, which degrades sequential read performance of the file system. In this paper, we analyze the relationship between file fragmentation and the sequential read performance, considering the characteristics of underlying storage devices. We also propose a novel file defragmentation scheme on LFS to effectively address the file fragmentation problem. Our scheme reorders the valid data blocks belonging to a victim segment based on the inode numbers during the cleaning process of LFS. In our experiments, our scheme eliminates file fragmentation by up to 98.5% when compared with traditional LFS.

URLhttps://dl.acm.org/doi/10.1145/2967360.2967622
DOI10.1145/2967360.2967622
Citation Keypark_file_2016