Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 24 Oct 1999 14:52:22 -0400 (EDT)
From:      Kevin Street <street@iname.com>
To:        freebsd-questions@FreeBSD.org
Cc:        Greg Lehey <grog@lemis.com>
Subject:   vinum stripe size vs ufs
Message-ID:  <14355.21862.534594.839643@mired.eh.local>

next in thread | raw e-mail | index | archive | help
I just set up a new vinum installation with a striped volume on 2
disks.  I noticed an ugly interaction between vinum's stripe sizes and 
ufs.  My first try used the default stripe size of 256k and newfs
defaults.  Watching iostat when I did the restore of my file system
to the new volume, I could see all the io going to one disk during the
first part of the restore (restore of the dir structure ?) then a 4 to 
1 imbalance of io during the rest of the restore.  A du on the
resulting file system did all its io on one disk.

I tried changing the newfs cylinders/group to 19 (from 16) but this
had little effect.  I then rebuilt my vinum plex with a stripe size of 
257k (rather than 256k) and newfs -c 19 and this works MUCH better.  I 
now see very even distribution of io during the restore, directory
operations and general file access.  Prime numbers == good.

I've probably managed to get vinum into a fairly optimal state by
tinkering, but are there any recommendations for stripe sizes or newfs 
parameters for vinum use on striped configurations?
 
Greg, you should probably think about changing the default stripe size 
since the default gives very non-optimal results on 2 disks (and
probably on any even number of striped disks).

This was all on a very recent 4.0 -current but probably applies to
vinum in 3.x as well.
-- 
Kevin Street
street@iname.com


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?14355.21862.534594.839643>