Dokáže ext2 efektivně pracovat s dlouhou partišnou?

Jan Kasprzak kas na informatics.muni.cz
Pátek Červenec 28 16:55:48 CEST 2000


Michal Samek wrote:
: V jedne z poslednich LinuxGazette je dost dobry clanek 
: popisujici principy a konretni implementace journal file
: systems pro (mimo jine) linux. Doporucuji procist.
: Dozvite se tam i kdy a proc ext2fs efektivni moc neni.
: Otazka je, zda 12GB je opravdu tak moc a o jake aplikace
: vam pujde.
: 
	Ve skutecnosti ten clanek moc dobry neni. Ext2 je pro vetsinu
veci pomerne dost efektivni. Neprilis caste pripady jako 10^4 souboru
v adresari pominu, ale pro vetsinu aplikaci opravdu ext2 staci.
Posilal jsem autorovi minuly tyden par pripominek:

	Hello Juan,

	I've read your article in Linux Gazette. There are few errors
in your article:

- in the first table: ext2fs/ext3fs file size is not limited to 2GB,
	it is around 17GB for a filesystem with 1024-bytes block size
	(256^3+256^2+256+10 1024-byte blocks).  The VFS limit, though
	lowers this to 2GB (but this is true for all other filesystems).
	For bigger block size the maximum file size is even larger.
- in the second table: ext2/ext3fs does support bitmap-management of the
	free blocks (and has even free inodes bitmap). It also supports
	the symbolic link in inode (up to 64 bytes of symlink text, I think),
	and newer versions of ext2fs even cache the file type in
	directory entry to speed up recursive searches (the namei() function).
- the third table: ext2fs/ext3fs indeed does support the sparse files
	(all filesystems with the same inode layout like ext2/ufs have
	do support this). On the other hand, it has to be stated that
	earlier implementation of XFS (up to IRIX 6.4, I think) did not
	support sparse files. You can verify this using the following
	commands:

		dd if=/etc/passwd of=sparse_file seek=1M
		ls -ls sparse_file

	(and then compare the file size with # of blocks used times block
	size). For me it led to the following output:

  12 -rw-r--r--    1 kas      staff    536872411 Jul 21 15:25 sparse_file

	(12 blocks - few data ones and some indirect blocks).


	There are some advanced features that you did not mention,
for example the bandwidth allocation of XFS, on-line resizing
of JFS, or space allocation of XFS (on XFS, the physical space
on disk is not allocated at the time of write(2), but rather at the
time of disk cache flush. This allows to merge several writes to one
physical extent, because at that time you know how many bytes you are
going to need).

	Other than that, I think your article is quite informative.
Thanks,

-Yenya

-- 
\ Jan "Yenya" Kasprzak <kas at fi.muni.cz>       http://www.fi.muni.cz/~kas/
\\ PGP: finger kas at aisa.fi.muni.cz   0D99A7FB206605D7 8B35FCDE05B18A5E //
\\\             Czech Linux Homepage:  http://www.linux.cz/              ///
/// Vite jak Microsoft vyrabi nezavirovana CD? ... ... ... Pouziva UNIX! \\\
//     http://support.microsoft.com/support/kb/articles/Q80/5/20.ASP      \\


Další informace o konferenci Linux