From owner-freebsd-current  Wed Nov  8 16:13:33 1995
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.6.12/8.6.6) id QAA00814
          for current-outgoing; Wed, 8 Nov 1995 16:13:33 -0800
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id QAA00808
          for <current@freebsd.org>; Wed, 8 Nov 1995 16:13:25 -0800
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id LAA16890; Thu, 9 Nov 1995 11:11:21 +1100
Date: Thu, 9 Nov 1995 11:11:21 +1100
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199511090011.LAA16890@godzilla.zeta.org.au>
To: current@freebsd.org, faulkner@mpd.tandem.com
Subject: Re: iozone and mount -o async
Sender: owner-current@freebsd.org
Precedence: bulk

>Can you test the fs mounted async with iozone?  I am getting a slowdown
>mounted async!

This is normal.  Mounting async disables the pessimization of doing
async (but ASAP) writes for full buffers.  This pessimization is
actually an optimization for stupid benchmarks such as iozone.  It
forces the physical i/o to be done ASAP instead of later when iozone
does an fsync().  Most applications don't do fsync()s so their i/o
gets done when the system thinks best.  It might be useful to have
an fadvise() syscall (like madvise() but working :-).  Applications
that do large sequential i/o's could use it to suggest write-through.

Bruce