Re: [Ext2-devel] disk throughput

Kai Henningsen (kaih@khms.westfalen.de)
06 Nov 2001 09:10:00 +0200


akpm@zip.com.au (Andrew Morton) wrote on 05.11.01 in <3BE77599.9CFB5CA9@zip.com.au>:

> Linus Torvalds wrote:
> >
> > I do believe that there are probably more "high-level" heuristics that can
> > be useful, though. Looking at man common big trees, the ownership issue is
> > one obvious clue. Sadly the trees obviously end up being _created_ without
> > owner information, and the chown/chgrp happens later, but it might still
> > be useable for some clues.

Size of the parent directory might be another clue.

> I didn't understand your objection to the heuristic "was the
> parent directory created within the past 30 seconds?". If the
> parent and child were created at the same time, chances are that
> they'll be accessed at the same time?

Thought experiment:

Put stuff on a disk the usual slow way.

Backup. Mkfs. Restore.

Should the allocation pattern now be different? Why?

> And there's always the `chattr' cop-out, to alter the allocation
> policy at a chosen point in the tree by administrative act.

Much help that's going to be in the above scenario, given how tar calls
chattr ... not.

> Any change in ext2 allocation policy at this point in time really,
> really worries me. If it screws up we'll have 10,000,000 linux
> boxes running like dead dogs in a year. So if we _do_ make a change, I'd
> suggest that it be opt-in. Call me a wimp.

Well, with Alex' cleanups, switchable policies might just become possible.

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