From owner-freebsd-hardware  Fri Sep 22 04:29:50 1995
Return-Path: owner-hardware
Received: (from root@localhost)
          by freefall.freebsd.org (8.6.12/8.6.6) id EAA26433
          for hardware-outgoing; Fri, 22 Sep 1995 04:29:50 -0700
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34])
          by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id EAA26423
          for <hardware@freebsd.org>; Fri, 22 Sep 1995 04:29:33 -0700
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id VAA03283; Fri, 22 Sep 1995 21:25:15 +1000
Date: Fri, 22 Sep 1995 21:25:15 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199509221125.VAA03283@godzilla.zeta.org.au>
To: cvance@cfar.umd.edu, hardware@freebsd.org
Subject: Re: Adaptech 1542 & 3Com 3c509 conflict?
Sender: owner-hardware@freebsd.org
Precedence: bulk

>I seem to have some sort of conflict between my network interface and
>scsi disks.  The scsi controller is an adaptech 1542CF and the eth
>adapter is a 3c509.

>Basically, the network is slow writing to a scsi disk, and fast the
>rest of the time.  Has anyone seen this before? Can anyone suggest
>something I could do to fix the situation?

Perhaps the scsi disk is inherently slower for the type of work given
to it by ftp.  This is not unlikely, since ftp's writes are poorly
buffered and scsi disks often have a very high command overhead.
ktracing ftp here showed the following sequence of write sizes:

..., 0x5a0, 0x10e0, 0x2760, 0xb40, 0x5a0, 0x10e0, 0x2760, 0x1c20, ...

The speed over ethernet (WD8013EBT) for ftp'ing a 15MB file to a slow
scsi disk and to /dev/null were the same (about 440K/sec, probably
limited by the slow IDE disk on the server).  This was faster than for
ftp'ing from localhost and the same speed as cp'ing from the same file
system in some cases (cases where the buffer cache cannot help and
where clustering is apparently ineffective).

Bruce