Faster Search Time Filesystem (FASTFS)

1
Faster Search Time Filesystem(FASTFS)

• Aaron Dingler and Steve Kurtz
• November 1 2007

2
Problem

• Exhaustively searching your filesystem for a word
  or filename is SLOW!
• All files must be searched
• Essentially your entire disk must be paged
  through memory
• We need to narrow our search while not losing
  breadth or accuracy

3
Solution

• Use FUSE to layer our solution on pre-existing
  filesystem
• Create a Library of key words and links to
  files/directories in which they exist
• But which key words do we cache? . . . Everything
• What is the best way to manage this Library? .
  . . A database

4
Implementation

• SQLite
• Quick access and modification
• Easy management

Word Word_id
file1 1
file2 2

Word_id Path_id
1 1
1 2
2 3
Path Path_id
/home/dir1/file1 1
/home/dir2/file1 2
/home/dir1/file2 3
5
Initial Results

• We have built our database and can insert, delete
  and modify files and directories
• Search file/directory names
• FUSE updates our database
• Average search time for find
• Standard .185s
• FASTFS .002s

6
Future Work

• Index words in files
• Measure FASTFS performance
• find grep
• Different filesystem structures
• Database space overheard
• Individual commands (write, mkdir, rm)
• Final results will be analyzing the effectiveness
  of FASTFS

7
Questions?