ADSM-L

Re: ADSM Database as raw device faster?

1998-07-27 15:03:14
Subject: Re: ADSM Database as raw device faster?
From: David Hendrix <dmhendri AT FEDEX DOT COM>
Date: Mon, 27 Jul 1998 13:03:14 -0600
My two cents:

We have multiple servers.  We prefer raw devices, that was suggested to
us two years ago.  However, HP and Sun do not support raw (at least that
I could find).  We use veritas and UFS filesystems for the adsm database
and log.  I have run multiple audits +fix on these systems because of
corruption.  Whether this was due to the ADSM code, the filesystem code
or what I am not sure.  I will say, the circumstance preceeding the
corruption was always a series of system crashes (for one reason or
another).

I have many other AIX systems.  One, running for 2+ years and sitting
around 10GB (with millions of entries) has never experienced corruption
despite me killing the machine quite a few times.

My experience leans towards raw, since I have had fewer problems using
raw.  On those systems where ADSM does not support raw, it's filesystems
or nothing.

David Hendrix
dmhendri AT fedex DOT com

Dwight Cook wrote:
>
>      In the past even IBM experts were split, I can't say how they feel now
>
>      What we've done is put the DB & LOG files under JFS and made all the
>      diskpool storage raw.
>
>      ALSO went to great lengths to put the volumes that contain a db or log
>      file as the first disk out a port (SSA) and that the db copies and log
>      copies were on a different card
>
> ______________________________ Reply Separator 
> _________________________________
> Subject: ADSM Database as raw device faster?
> Author:  g.rentschler (g.rentschler AT rus.uni-stuttgart DOT de) at unix,mime
> Date:    7/27/98 10:04 AM
>
> Hi,
> we have a RS/6000 as ADSM server. So far I created jfs filesystem for the
> database and created the database volumes as normale files via dsmfmt. As
> far as I know it is possible to use a logical volume as raw device for this
> purpose. What is the recommended way? Is the raw device faster?
> Best regards
> Gerhard
>
> Gerhard Rentschler            email: g.rentschler AT rus.uni-stuttgart DOT de
> Regional Computing Center     tel.   ++49/711/685 5806
> University of Stuttgart       fax:   ++49/711/682357
> Allmandring 30a
> D 70550 Stuttgart
> Germany