Re: [Ext2-devel] disk throughput

[Linus Torvalds]

On Mon, 5 Nov 2001, Andrew Morton wrote:
>
> 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?

the thing I don't like about it is the non-data-dependence, ie the
layout of the disk will actually depend on how long it took you to write
the tree.

I'm not saying it's a bad heuristic - it's probably a fine (and certainly
simple) one. But the thought that when the NFS server has problems, a
straight "cp -a" of the same tree results in different layout just because
the server was moved over from one network to another makes me go "Ewww.."

		Linus