From owner-freebsd-hackers  Thu Apr 18 13:55:31 1996
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id NAA08298
          for hackers-outgoing; Thu, 18 Apr 1996 13:55:31 -0700 (PDT)
Received: from localhost.cdrom.com (localhost.cdrom.com [127.0.0.1])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id NAA08285
          Thu, 18 Apr 1996 13:55:27 -0700 (PDT)
Message-Id: <199604182055.NAA08285@freefall.freebsd.org>
X-Authentication-Warning: freefall.freebsd.org: Host localhost.cdrom.com [127.0.0.1] didn't use HELO protocol
To: Jaye Mathisen <mrcpu@cdsnet.net>
cc: hackers@freebsd.org, gibbs@freebsd.org
Subject: Re: QUEUE_FULL_ENABLE option really work? 
In-reply-to: Your message of "Thu, 18 Apr 1996 13:01:06 PDT."
             <Pine.NEB.3.92.960418125843.19574E-100000@schizo.cdsnet.net> 
Date: Thu, 18 Apr 1996 13:55:26 -0700
From: "Justin T. Gibbs" <gibbs@freefall.freebsd.org>
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

The option is QUEUE_FULL_SUPPORTED and it does what its supposed to.
It increases the number of tags allowed per device to 4 instead of
the default of two.

>I just turned it off on a box with a 2940, and across the board I'm
>picking up 700-800k/s improvements:
>
>old:
>
>IOZONE performance measurements:
>        1168024 bytes/second for writing the file
>        4445767 bytes/second for reading the file
>
>
>new:
>
>IOZONE performance measurements:
>        1790285 bytes/second for writing the file
>        5332448 bytes/second for reading the file

But your random I/O scores will decrease since the drive will only
have at max two I/Os to sort in order to reduce seeks.  If you're
only interested in sequential I/O, you might as well turn off
tagged queueing since for some devices you will get better results.

--
Justin T. Gibbs
===========================================
  FreeBSD: Turning PCs into workstations
===========================================