[2511] in linux-scsi channel archive

home help back first fref pref prev next nref lref last post

Re: Adaptec 2940 support

daemon@ATHENA.MIT.EDU (Zlatko Calusic)
Mon Sep 22 13:42:19 1997

To: Jeff Noxon <jeff@planetfall.com>
Cc: linux-scsi@vger.rutgers.edu
Reply-To: Zlatko.Calusic@CARNet.hr
From: Zlatko Calusic <Zlatko.Calusic@CARNet.hr>
Date: 	22 Sep 1997 18:52:30 +0200
In-Reply-To: Jeff Noxon's message of Mon, 22 Sep 1997 11:13:50 -0500

Jeff Noxon <jeff@planetfall.com> writes:

> The aic7xxx chips are mostly fine from a technical standpoint, but the
> Linux driver (or any free driver for that matter) is not without problems.
> Adaptec is very restrictive with programming information for these chips.

Yes, I heard about that, and find it very annoying if not bad company
policy. I decided to go with Adaptec, mainly because it's easy to get
it here, which is not true of other SCSI controller types (BusLogic
for example).

> The driver is not good at recovering from error conditions, it has many
> bugs, and it could probably be faster.
>

Well, this is what concerns me the most. I hope that I didn't throw
lots of money on something that will in the end prove as unstable or
of bad performance. Of course, Linux driver quality is of the biggest
importance here, and I hope it will not fail in accomplishing it's
task. :)

> For what it's worth, I have a 3940W on one of my servers and it works
> fine.  The driver is still getting better every day.
>

As I can see, too, driver is improving constantly. I'm just hoping it's 
not because it's unstable. :)

> Still, given Adaptec's attitude toward Linux, I will not buy any of their
> products given the choice.  BusLogic supports Linux hand has a fantastic,
> fast, robust driver.  Same goes for DPT if you need a high-end controller.
> 
> > I have one in the PCI slot (to accommodate CD writing), and am planning 
> > to attach Sun's SCSI disk on it.
> > 
> > Does anybody have experiences of such configuration in work?
> > Especially performance wise.
> 
> It will work.  On my 3940W with a raid 0 of two IBM wide 2.2gb 7200rpm
> disks, I get about 9MB/sec.  If I read one disk, I get about 7MB/sec.
> This is with dd and a 64k blocksize.  I haven't tried these disks with
> a BusLogic or DPT controller yet, so I can't comment on how good or bad
> the raw throughput is.

7/9 MB/sec looks good to me.

My IBM-DAQA33240 (3.2GB EIDE) performs at about 4.5 MB/sec, which I
find fast enough, but I'm not satisfied with CPU consumption, of
course. I believe that the biggest difference between my Pentium Linux
(133Mhz) and UltraSparc Solaris (143Mhz) is in quality of I/O
subsystem (SCSI <-> EIDE).

I hope that I'm right, and that connecting SCSI disk will greatly
improve perception of speed on my (beloved :) ) Linux machine.

Could you (or anybody else) send some 'bonnie' benchmarks of SCSI
disks for me to check CPU usage in such systems?

This is what I got using Linux 2.1.56, 100MB file, Pentium 133MHz,
32MB RAM, (already mentioned) EIDE IBM 3.2GB:

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
          100  3819 90.3  6100 51.4  1762 39.8  2821 82.7  4860 69.6  49.6  4.4

-- 
Posted by Zlatko Calusic           E-mail: <Zlatko.Calusic@CARNet.hr>
---------------------------------------------------------------------
	       Surprise your boss. Get to work on time.

home help back first fref pref prev next nref lref last post