On Thu, 15 May 2003 16:13:14 +0100
"Wilcox, Andy" <andy.wilcox AT AQUILA-NETWORKS.CO DOT UK> wrote:

> Use dsmfmt to create your log and DB volumes and then use dsmserv
> format to format the recovery log space and create the dsmserv.dsk.
> Please note that dsmserv.dsk will be created in the current directory.
>
> eg
>
> cd /tsm/server0 #where this $DSMSERV_CONFIG directory
> dsmfmt -m -db /tsm/server0/db/dbvol1 1000 #where the last number is
> size of volume in MB
> dsmfmt -m -db /tsm/server0/db/dbvol2 1000

Which works, but when tweeking for performance on our TSM server, I got
the very strong impression that TSM tries to do smart things with
volumes.

It assumes that if you have multiple db volumes, it gives better
performance to 'stripe' across those than first filling one and than the
other. Which means that in your example you have a slightly decreased
performance that just defining one 2000 MB large DBvol, and a sligtly
bigger change of trashing effects on the disk. I've stuck to
one dbvol per phisical volume, and it has served me well (remember me
bragging about a 5.1 upgrade with just 15 minutes of upgradedb on our
100 GB db just yesterday?)

> dsmfmt -m -db /tsm/server0/db/logvol1 500
>
> dsmserv format 1 /tsm/server0/db/logvol1 2 /tsm/server0/db/dbvol1
> /tsm/server0/db/dbvol2 #where 1 is the number of following log file
> specifications and 2 is the number of following db file specs.
>
> hope this helps a bit more...
>
> Cheers
>
> Andy Wilcox
> Midrange Services
> Aquila Networks Services Ltd
>

--

Remco