lkml.org 
[lkml]   [1999]   [May]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: Deletion of big files...
Date
From
Alexander Viro wrote:
>
>
> On Wed, 19 May 1999, Rogier Wolff wrote:
>
> > Alexander Viro wrote:
> > >
> > >
> > > On Tue, 18 May 1999, Rogier Wolff wrote:
> > >
> > > > Guys, this is all convoluted ways of fighting symptoms. The cause of
> > > > "rm taking a long time" is that the indirect blocks are scattered all
> > > > over the disk.
> > >
> > > Nope. It is $DEITY-awful amount of waiting/retries/schedule() calls in
> > > ext2_truncate(). Look and you'll see. That, and zeroing indirect blocks
> > > (unnecessary for this usage of ext2_truncate()).
> >
> > Ok, So we could speed things up a bit by not writing back the indirect
> > blocks. Under minix-fs, datablocks that got freed before getting
> > written to disk (e.g. a tempfile from the compiler), would never get
> > written. I've been told that ext2fs does the same. Doesn't the same
> > hold for the indirectblocks? They should get freed after getting
> > zeroed, and then never getten written to disk.
>
> Sorry, no. Doesn't work that way.

Something to fix then.....

> > The "improve performance by a factor of 16" which i predicted for a 4k
> > blocksize filesystem was a factor fo 30 according to someones (pretty
> > inaccurate) measurements.

Oh. I forgot to mention, mke2fs is going to default to 4k blocksize in
the future for filesystems that are large enough to hold "large"
files. That will make all this less annoying, as most people will
just use the default.

> IOW complexity of current implementation is pretty bad. And that can be
> fixed. Putting indirect blocks together may improve performance on
> truncate() (after all, disks usually read the whole track and have cache),
> but will hamper sequential reads.

... about 30 ms per 256k read, provided you screw up the read-ahead of
the indirect blocsk.

Roger.


--
** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2137555 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
------ Microsoft SELLS you Windows, Linux GIVES you the whole house ------


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:51    [W:0.031 / U:1.220 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site