Re: [Ext2-devel] disk throughput

[Jeremy Fitzhardinge <jeremy@goop.org>]

[-- Attachment #1: Type: text/plain, Size: 1162 bytes --]

On Tue, 2001-11-06 at 02:52, Daniel Phillips wrote:
> On November 5, 2001 11:22 pm, Andrew Morton wrote:
> > - The seek distance-versus-cost equation has changed.  Take a look
> >   at a graph of seek distance versus time.  Once you've decided to
> >   seek ten percent of the distance across the disk, a 90% seek only
> >   takes twice as long.
> 
> Do you have such a graph handy?

This is one I made a while ago while doing some measurements; its also
the one Andrew was referring to.  The drive is a very ordinary WD 40G
7200RPM drive.  You can clearly see the rotation time in the width of
the band, and the non-linear short-seek optimizations (the drive seems
to do something like dump a lot more current into the seek coil for
short seeks, probably knowing that it can't burn things out if its only
short).

There's an interesting detail you can't really see from this graph. 
Very short seeks (<< 1% disk size) are about 1ms.  That's not seeks
within a track; that's track-track seeks.  Since rotational latency is
up to 8.3ms on this drive, its often cheaper to seek to the next track
if the data is available there rather than wait for a rotation.

	J

[-- Attachment #2: seek-buf.eps.gz --]
[-- Type: application/x-gzip, Size: 26186 bytes --]