On Tue, 26 Dec 1995 18:28:59 +0100 John O'Neall said:
>We're installing the ADSM/6000 server on AIX 3.2.5.  One can use either
>files created by DSMFMT or raw logical volumes created by smit for
>random-access storage-pool volumes.  Has anyone compared these two
>methods?  Is there any reason for using one rather than the other?

It's harder to inadvertently screw up a JFS file.      If you use
a raw logical volume, and if you increase its size (e.g., using SMIT), then
you can pretty easily screw up the volume.  However, I believe that raw LV's
are more efficient than JFS volumes.  This belief is based on the fact that
ADSM provides it's own mirroring and buffer caching (for DB and LOG vols),
and therefore the JFS overhead to do this is redundant and wasteful of system
CPU and memory resource.  For storage volumes, if you want mirroring, then
that is a reason to use JFS since ADSM does not provide mirroring for disk
storage volumes.  However, if you don't care about mirroring, I think raw LV's
are again better than JFS.  While ADSM does not provide buffer cacheing for
disk storage volumes, I don't think that is much of an issue, since most data
in disk storage volumes is not repetitively accessed.  I.e., you store the
backup/archive data on disk and normally don't access it again immediately.

..Paul

Paul Zarnowski                     Phone:   607/255-4757
Cornell Information Technologies   Fax:     607/255-6523
Cornell University                 US Mail: 315 CCC, Ithaca, NY 14853-2601