[2515] in linux-scsi channel archive
Re: Adaptec 2940 support
daemon@ATHENA.MIT.EDU (Jeff Noxon)
Mon Sep 22 16:33:15 1997
Date: Mon, 22 Sep 1997 14:35:27 -0500
From: Jeff Noxon <jeff@planetfall.com>
To: Zlatko.Calusic@CARNet.hr
Cc: Jeff Noxon <jeff@planetfall.com>,
David Sparc Miller <davem@jenolan.rutgers.edu>,
linux-scsi@vger.rutgers.edu
In-Reply-To: <87lo0pw89a.fsf@atlas.infra.CARNet.hr>; from Zlatko Calusic on Mon, Sep 22, 1997 at 07:47:29PM +0200
On Mon, Sep 22, 1997 at 07:47:29PM +0200, Zlatko Calusic wrote:
> I decided to lower the reset delay to 5 seconds (from the default 15)
> because my kernel completely locks while SCSI devices are being
> reseted, and I don't like that behaviour. Now my Linux is back sooner,
> atleast. :)
I tried that at first, but now I use this in lilo.conf:
append="aic7xxx=no_reset"
The reset doesn't seem to do anything -- the BIOS resets the card anyway.
> Hm, I can't say that those %CPU columns are low.
> Compared to results I got on Sun, these are definitely too high.
Maybe, maybe not. Comparing your Sun numbers between Linux and Slowaris
on the same box, I see lower throughput and lower CPU utilization. It seems
fair that you'd get lower utilization when you're moving less data.
My numbers were generated on a slow machine, so it's obviously going to
take a greater percent of the CPU's muscle. I have a P120 w/ 256K pburst,
32M 70ns FPM memory, and the old FX chipset. If I tried the same test on
a PPro with the same software configuration, I'd expect the utilization
numbers to be _much_ lower. Unfortunately I don't have any fast boxes
to try it on. (I wouldn't be surprised, though, if my box ran faster
I/O at 100 MHz. Maybe I'll try that.)
I still don't know why R/W performance doesn't scale well with two drives
and adapters.
Maybe a guru can tell us for sure what is happening. Dave? :)
Thanks,
Jeff
