Skip to Content.
Sympa Menu

sm-discuss - Re: [SM-Discuss] reiserfs and noatime

sm-discuss AT lists.ibiblio.org

Subject: Public SourceMage Discussion List

List archive

Chronological Thread  
  • From: Seth Woolley <seth AT tautology.org>
  • To: Robin <robin AT kallisti.2y.net>
  • Cc: sm-discuss AT lists.ibiblio.org
  • Subject: Re: [SM-Discuss] reiserfs and noatime
  • Date: Mon, 30 Jun 2003 17:12:20 -0700 (PDT)

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I would think smaller trees would be an advantage for writes.
tails give some internal fragmentation benefit, but:

http://www.informatik.uni-frankfurt.de/~loizides/reiserfs/
http://www.informatik.uni-frankfurt.de/~loizides/reiserfs/oldpage/gif/4G-cpu-nodel.gif

that image shows how much more tails requires more cpu usage.

At least for writes, it says:
"Omitting the fragments by mounting with "-notail" improves the
performance (absolutely and in shape)"

Moving around bigger trees must be a performance hit.

I suspect reads would go faster with tails for the reasons you stated, and
I suppose if your goal is to do lots of reading but not much writing,
tails would be a good option.

This distro though, does lots of writing... hmm, but mostly
the killer on my laptop's is decompressing tarballs, but 2.4.21 really
helped there.

In the end, I'd be curious as to compare a gcc compile with and without
tails.

I bet disk cache would make the difference slight.

Seth

On Tue, 1 Jul 2003, Robin wrote:

> On Mon, Jun 30, 2003 at 11:38:33AM -0700, Seth Woolley wrote:
> > notail is more speedy, but uses more space.
> > noatime gets a little performance benefit, but by default, I think we
> > should enable atime or ask during install?
> Yeah, thats my thought also. Especially as it messes with some things.
> I'm curious about notail being faster, I would have expected it to be a
> little slower, as if a file is packed into the directory tree, then
> there is less seeking to find it, and a greater chance of it being
> cached, compared to it being with the normal files.
>
> > I don't know about the confusing lilo bit.
> That was from the mount(8) man page when I was looking it up:
> notail By default, reiserfs stores small files and `file
> tails'
> directly into its tree. This confuses some utilities such
> as
> LILO(8). This option is used to disable packing of files
> into
> the tree.
>
> --
> Robin <robin AT kallisti.2y.net> JabberID: <eythian AT jabber.org>
>
> Hostes alienigeni me abduxerunt. Qui annus est?
>
> PGP Key 0x776DB663 Fingerprint=DD10 5C62 1E29 A385 9866 0853 CD38 E07A 776D
> B663
> NB: New key! Available from http://kallisti.2y.net/~robin/newgpgkey
>

- --
Seth Alan Woolley <seth at tautology.org>, SPAM/UCE is unauthorized
Key id 7BEACC7D = 2978 0BD1 BA48 B671 C1EB 93F7 EDF4 3CDF 7BEA CC7D
Full Key at seth.tautology.org and pgp.mit.edu. info: www.gnupg.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.0 (FreeBSD)

iD8DBQE/ANHm7fQ833vqzH0RAsNvAJ42wKuPbYk5twF0JbqeONbwIoWhjACg2nuI
hke2O4UC4aIZyVFZMUcI9EM=
=ZjZP
-----END PGP SIGNATURE-----





Archive powered by MHonArc 2.6.24.

Top of Page