On Mon, 18 Oct 2004, Manuel Bouyer wrote:

> Date: Mon, 18 Oct 2004 23:58:47 +0200
> From: Manuel Bouyer <bouyer@antioche.eu.org>
> To: Karl Janmar <karl@utopiafoundation.org>
> Cc: tech-kern@NetBSD.org
> Subject: Re: ATI IXP IDE chipset support
>
> On Mon, Oct 18, 2004 at 11:54:27PM +0200, Karl Janmar wrote:
> > I did the read test with dd now, first making sure that the file wasn't in
> > the cache.
> >
> > # dd of=/dev/null if=apslaskr bs=4096 count=100000
> > sync100000+0 records in
> > 100000+0 records out
> > 409600000 bytes transferred in 14.996 secs (27313950 bytes/sec)
> >
> > I could also se on sysstat vmstat that I hade about 25M of disktransfer
> > from wd0, as expected. So it seems that there is no problem to read/write
> > to the blockdevice on my disk... but the character device....
>
> This is strange, my test program does 64k reads, just as the block device does.
> But my test program doens't do a sequential read, it always read the
> same 64k over and over. Also it starts from sector 0.
> Maybe this is the difference. Could you try a dd from the char device to
> /dev/null with bs=64k ?
>

Okey, strange things happening, I ran your program again and got
the following result:

the ATA disk:
# ./tst /dev/rwd0d 30000
25105444 us, 74.684997 MB/s

the SATA disk:
# ./tst /dev/rwd1d 30000
48555801 us, 38.615365 MB/s

So now I got the expected results.


I don't exacly know the reason for this to "just" work now, It has a RAID
configured for both disks but I havn't used any filsystem on that for
doing these tests ofcourse (used a partition on one of the disks that are
bigger).

The system didn't do any reconstruction or anything when I did the tests in the
beginning, and the strange thing here IF something with the idleing RAID
configuration did the performance degree, why was it only on the raw
device?


- Karl