From owner-freebsd-hackers  Sat Mar  8 23:31:37 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id XAA04872
          for hackers-outgoing; Sat, 8 Mar 1997 23:31:37 -0800 (PST)
Received: from scanner.worldgate.com (scanner.worldgate.com [198.161.84.3])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id XAA04861;
          Sat, 8 Mar 1997 23:31:32 -0800 (PST)
Received: from znep.com (uucp@localhost) by scanner.worldgate.com (8.8.5/8.7.3) with UUCP id AAA08370; Sun, 9 Mar 1997 00:31:23 -0700 (MST)
Received: from localhost (marcs@localhost) by alive.znep.com (8.7.5/8.7.3) with SMTP id AAA09446; Sun, 9 Mar 1997 00:31:39 -0700 (MST)
Date: Sun, 9 Mar 1997 00:31:38 -0700 (MST)
From: Marc Slemko <marcs@znep.com>
To: Brian Tao <taob@risc.org>
cc: "matthew c. mead" <mmead@goof.com>, isp@freebsd.org, hackers@freebsd.org
Subject: Re: freebsd as a news server?
In-Reply-To: <Pine.BSF.3.95.970309021448.22711C-100000@alpha.risc.org>
Message-ID: <Pine.BSF.3.95.970309002758.6158R-100000@alive.znep.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

You are probably using 2.2 or -current, right?  On a 2.2 system I get
similar results to yours.  On 2.1, async mounts only change one bit of
ffs code.  In 2.2, they make more things async.  I don't think the
difference in real life between 2.1 async and 2.2 async is as big as in a
test like this.

On Sun, 9 Mar 1997, Brian Tao wrote:

> On Sat, 8 Mar 1997, Marc Slemko wrote:
> > 
> > #!/usr/bin/perl
> > for ($i = 0; $i < 1000; $i++) {
> >         open(F, ">test.$i");
> >         close F;
> > }
> > for ($i = 0; $i < 1000; $i++) {
> >         unlink("test.$i");
> > }
> >
> > This gives me 36.62 seconds sync and a drive that sounds like it is
> > trying to chew on a pengiun vs. 22.99 async and an almost silent
> > drive.
> 
>     I was quite sure the difference was much greater, so I tried it on
> my system at home after seeing your numbers:
> 
> # mount -u -o async,noatime /
> # cd /tmp ; time touch `jot 1000` ; time rm `jot 1000`
> 0.055u 1.452s 0:01.55 96.7% 17+186k 1+24io 0pf+0w
> 0.062u 0.371s 0:01.10 39.0% 175+244k 0+23io 0pf+0w
> # sync
> # mount -u /
> # time touch `jot 1000` ; time rm `jot 1000`
> 0.062u 1.655s 0:34.51 4.9% 16+183k 0+2023io 0pf+0w
> 0.047u 0.618s 0:30.19 2.1% 178+242k 0+2000io 0pf+0w
> 
>     2.65s vs. 64.70s in tcsh, and 1.72s vs. 44.44s using your perl
> example.  Why the large discrepancy in async times, I wonder?
> --
> Brian Tao (BT300, taob@risc.org)
> "Though this be madness, yet there is method in't"
> 
>