Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 15 Mar 2001 20:51:36 -0600
From:      David Kelly <dkelly@hiwaay.net>
To:        javor.evstatiev@blue-c.com
Cc:        freebsd-questions@FreeBSD.ORG
Subject:   Re: poor scsi hd performance 
Message-ID:  <200103160251.f2G2pae93330@grumpy.dyndns.org>
In-Reply-To: Message from javor.evstatiev@blue-c.com  of "Thu, 15 Mar 2001 20:15:54 %2B0100." <OF9F911B27.19172ED9-ONC1256A10.006909CA@blue-c.com> 

next in thread | previous in thread | raw e-mail | index | archive | help
javor.evstatiev@blue-c.com writes:
> bash-2.04# time tar xzf apache_1.3.19.tar.gz
> 
> real    0m8.723s
> user    0m0.267s
> sys     0m0.216s
> 
> I can't believe it should take 9 secs to decompress a 2 meg tarfile :( All
> linux boxes are faster - even the IDE based desktops...
> 
> Any ideas what/how to tune ? Thx 4 any info in advance!

For a more realistic measure of SCSI thruput play with bonnie, available
in ports. Make sure your test file is at least twice the size of your
RAM.

Took my 800 MHz Athlon and FW SCSI this amount of time:
0.271u 0.165s 0:10.03 4.2%      205+412k 51+22io 0pf+0w

And again a moment later:
0.242u 0.189s 0:06.97 6.0%      208+460k 0+24io 0pf+0w

Almost no time to remove (I have softupdates enabled)
% time rm -rf apache_1.3.19/
0.000u 0.026s 0:00.02 100.0%    156+204k 0+0io 0pf+0w

And writing to a dangerous write-caching-enabled ATA-100 drive (reading 
from SCSI, and by now the system's cache):
0.234u 0.180s 0:04.43 9.2%      217+495k 106+6io 0pf+0w

There are a lot of files and directories in the apache distribution. BSD
consciously made the decision to be no faster than the designers could
protect the filesystem consistency. File and directory creation are
possibly the worst case path thru the filesystem layer.

% tar -xvzf /usr/ports/distfiles/apache_1.3.19.tar.gz | wc -l
     805

% du -sk .
7397    .

May have been only 1.9MB to begin with but over 7MB after. 2MB of 
reading plus 7MB of writing in say 9 seconds is 1 MB/sec thruput with 
lots of seeks. Its not bad at all.

If you want to play with tweaking your system then you could use
camcontrol to set WCE in mode page 8 of your SCSI drive. Not recommended
if your data is precious. Bit may already be set. If it were an SGI or
Sun OEM disk then it would not be set.

Also, "options SOFTUPDATES" is in the GENERIC kernel. So you could boot
single user and use tunefs to enable SOFTUPDATES on your filesystems and
see what that does for you. Don't believe it ever hurts but doesn't
always help.

You can tell if softupdates is active by typing "mount"

% mount
/dev/da0s2a on / (ufs, local, soft-updates)
/dev/da0s2f on /usr (ufs, local, soft-updates)
/dev/da0s2e on /var (ufs, local, soft-updates)
/dev/ad4s3e on /usr2 (ufs, local, soft-updates)
/dev/ad4s3f on /usr3 (ufs, local, soft-updates)
/dev/ad4s3g on /usr4 (ufs, local, soft-updates)



--
David Kelly N4HHE, dkelly@hiwaay.net
=====================================================================
The human mind ordinarily operates at only ten percent of its
capacity -- the rest is overhead for the operating system.



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200103160251.f2G2pae93330>