Memory requirements

Richard W.M. Jones rjones at redhat.com
Wed Aug 31 08:47:46 UTC 2011


On Tue, Aug 30, 2011 at 03:36:52PM -0500, Bruno Wolff III wrote:
> On Tue, Aug 30, 2011 at 13:33:06 -0700,
>   Jeremiah Summers <jmiahman at gmail.com> wrote:
> > 
> > Yes Literally I did, but as Adam just pointed out running Live just
> > dumps the squashfs image to the drive and slaps grud on it. I'm not
> 
> It actually dumps the ext4 image on the drive and then resizes it to
> fit the available space. The ext4 image is stored compressed inside
> of a squashfs file system on the live image.
> 
> It would be nice to get rid of the embedded ext4 image now that squashfs
> supports special files and extended attributes (needed for selinuix labels),
> but there are some other roadblocks that will block that change for the near
> future.

There are a couple of issues here:

(1) Our current method of resizing the ext4 filesystem to minimal size
and then expanding it on the target mangles the ext4 filesystem layout
and apparently makes eventual performance poor.  I don't exactly
understand the details of what is happening but Ric Wheeler (CC'd)
should do.

(2) Expanding an ext4 filesystem is really fast because we're
bypassing the whole VFS layer when doing this.  That's where the
benefit of this method comes from.  I somehow doubt that unpacking a
squashfs will be nearly as fast (in my tests, unpacking a filesystem
from a tarball is 10x slower than just resizing a filesystem).  So the
main benefit of the live CD technique -- speed -- will be lost if we
make this change.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming blog: http://rwmj.wordpress.com
Fedora now supports 80 OCaml packages (the OPEN alternative to F#)
http://cocan.org/getting_started_with_ocaml_on_red_hat_and_fedora


More information about the devel mailing list