From: Andreas Dilger <adilger@turbolinux.com>
To: Daniel Phillips <phillips@innominate.de>
Cc: "H. Peter Anvin" <hpa@transmeta.com>,
Linux-Kernel <linux-kernel@vger.kernel.org>
Subject: Re: [rfc] Near-constant time directory index for Ext2
Date: Wed, 21 Feb 2001 19:03:07 -0700 (MST) [thread overview]
Message-ID: <200102220203.f1M237Z20870@webber.adilger.net> (raw)
In-Reply-To: <3A9469D8.DB4679DB@innominate.de> from Daniel Phillips at "Feb 22, 2001 02:22:32 am"
Daniel Phillips writes:
> Easy, with average dirent reclen of 16 bytes each directory leaf block
> can holds up to 256 entries. Each index block indexes 512 directory
> blocks and the root indexes 511 index blocks. Assuming the leaves are
> on average 75% full this gives:
>
> (4096 / 16) * 512 * 511 * .75 = 50,233,344
>
> I practice I'm getting a little more than 90,000 entries indexed by a
> *single* index block (the root) so I'm not just making this up.
I was just doing the math for 1k ext2 filesystems, and the numbers aren't
nearly as nice. We get:
(1024 / 16) * 127 * .75 = 6096 # 1 level
(1024 / 16) * 128 * 127 * .75 = 780288 # 2 levels
Basically (IMHO) we will not really get any noticable benefit with 1 level
index blocks for a 1k filesystem - my estimates at least are that the break
even point is about 5k files. We _should_ be OK with 780k files in a single
directory for a while. Looks like we will need 2-level indexes sooner than
you would think though. Note that tests on my workstation showed an average
filename length of 10 characters (excluding MP3s at 78 characters), so this
would give 20-byte (or 88-byte) dirents for ext3, reducing the files count
to 4857 and 621792 (or 78183 and 40029696 for 4k filesystems) at 75% full.
Cheers, Andreas
--
Andreas Dilger \ "If a man ate a pound of pasta and a pound of antipasto,
\ would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/ -- Dogbert
next prev parent reply other threads:[~2001-02-22 2:04 UTC|newest]
Thread overview: 80+ messages / expand[flat|nested] mbox.gz Atom feed top
2001-02-20 15:04 Daniel Phillips
2001-02-20 20:03 ` Linus Torvalds
2001-02-20 21:08 ` Jeremy Jackson
2001-02-20 21:20 ` Mike Dresser
2001-02-20 22:36 ` Jeremy Jackson
2001-02-20 23:08 ` Daniel Phillips
2001-02-21 1:04 ` Bernd Eckenfels
2001-02-21 16:38 ` Daniel Phillips
2001-02-20 22:58 ` Jonathan Morton
2001-02-20 21:41 ` Daniel Phillips
2001-02-21 0:22 ` Linus Torvalds
2001-02-21 0:30 ` Alan Cox
2001-02-21 2:35 ` Ed Tomlinson
2001-02-21 23:13 ` Linus Torvalds
2001-02-21 23:34 ` Davide Libenzi
2001-02-21 23:59 ` Linus Torvalds
2001-02-21 23:57 ` H. Peter Anvin
2001-02-22 0:35 ` Ed Tomlinson
2001-02-21 1:01 ` Andreas Dilger
2001-02-22 2:28 ` Daniel Phillips
2001-02-22 3:30 ` Linus Torvalds
2001-02-22 16:33 ` Chris Mason
2001-02-22 22:30 ` Daniel Phillips
2001-02-21 17:21 ` Davide Libenzi
2001-02-21 21:08 ` Martin Mares
2001-02-21 21:29 ` Davide Libenzi
2001-02-21 21:32 ` Martin Mares
2001-02-21 21:59 ` Davide Libenzi
2001-02-21 22:26 ` Martin Mares
2001-02-21 22:43 ` Davide Libenzi
2001-02-21 22:14 ` H. Peter Anvin
2001-02-21 22:32 ` Martin Mares
2001-02-21 22:38 ` H. Peter Anvin
2001-02-21 22:50 ` Martin Mares
2001-02-21 22:54 ` H. Peter Anvin
2001-02-21 23:07 ` Martin Mares
2001-02-21 23:15 ` H. Peter Anvin
2001-02-21 23:42 ` Daniel Phillips
2001-02-21 23:52 ` Davide Libenzi
[not found] ` <3A945081.E6EB78F4@innominate.de>
2001-02-21 23:48 ` H. Peter Anvin
2001-02-22 1:22 ` Daniel Phillips
2001-02-22 1:42 ` H. Peter Anvin
2001-02-22 2:03 ` Andreas Dilger [this message]
2001-02-22 2:41 ` H. Peter Anvin
2001-02-22 3:43 ` Daniel Phillips
2001-02-22 4:02 ` Linus Torvalds
2001-02-22 5:19 ` Linus Torvalds
2001-02-22 11:31 ` Ingo Oeser
2001-02-22 18:20 ` Linus Torvalds
2001-02-22 4:02 ` H. Peter Anvin
2001-02-22 7:03 ` Andreas Dilger
2001-02-22 4:03 ` H. Peter Anvin
2001-02-22 10:35 ` Alan Cox
2001-02-23 0:59 ` Felix von Leitner
2001-02-22 3:08 ` Daniel Phillips
2001-02-22 8:06 ` [rfc] [LONG] " Andreas Dilger
2001-02-22 7:20 ` [rfc] " Bill Wendling
2001-02-22 8:34 ` Rogier Wolff
2001-02-21 23:26 ` Jamie Lokier
2001-02-22 19:04 ` Kai Henningsen
2001-02-22 6:23 ` [Ext2-devel] " tytso
2001-02-22 7:24 ` Daniel Phillips
2001-02-22 13:20 ` tytso
2001-02-22 18:16 ` Andreas Dilger
2001-02-22 23:04 ` Daniel Phillips
2001-02-22 23:40 ` tytso
2001-02-23 20:11 ` tytso
2001-02-24 0:32 ` Andreas Dilger
2001-02-22 18:38 ` Kai Henningsen
[not found] <Pine.LNX.4.10.10102211740550.1933-100000@coffee.psychology.mcmaster.ca>
2001-02-21 22:47 ` H. Peter Anvin
2001-02-23 1:52 Andries.Brouwer
2001-02-23 21:43 ` Ralph Loader
2001-02-23 22:37 ` Guest section DW
2001-02-24 2:47 ` Ralph Loader
2001-02-24 5:34 ` Ralph Loader
2001-02-23 2:49 Andries.Brouwer
2001-02-23 3:42 ` Daniel Phillips
2001-02-23 12:20 ` Jonathan Morton
2001-02-23 18:57 ` Andreas Dilger
2001-02-23 12:38 Andries.Brouwer
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=200102220203.f1M237Z20870@webber.adilger.net \
--to=adilger@turbolinux.com \
--cc=hpa@transmeta.com \
--cc=linux-kernel@vger.kernel.org \
--cc=phillips@innominate.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®