Jim,

fixediobufsize was a disaster since it corrupted backups.  Tivoli
has completely disabled it, and it isn't likely to return from what I have
been able to learn at SHARE.

The reason for dropping tapeiobufs when running on r10 to 3590's is
that tsm will write a 256k block.  It is still using bsam but it is exploiting
the 'large blocksize support' added to dfsms 1.6 which is an element of r10.

Also available in r10 is striping for lds files.  I am in the process of
converting all the tsm lds files to striped lds files to improve performance.

I am running 4.2.1 and it seems to be a very good level on os/390; none of
the problems seen by aix & nt servers apply to the os/390 version.  You need
4.2.* to get the 256k blocks on 3590 feature.

Hope this helps,

--
--------------------------
--------------------------
Bill Colwell
Bill Colwell
C. S. Draper Lab
Cambridge, Ma.
bcolwell AT draper DOT com
--------------------------
In <3BE16F62.5924A1AB AT email.unc DOT edu>, on 11/01/01
In <3BE16F62.5924A1AB AT email.unc DOT edu>, on 11/01/01
   at 11:14 AM, Jim Kirkman <jmk AT EMAIL.UNC DOT EDU> said:

>Hey,

>A search of the archives on this topic found some posts from back in Jan
>about how this option was broke in 3.7.sumthing, Bill Colwell was one of
>the respondees.

>Anyone been brave enough to experiment with it in v4 (we're at 4.1.4)?
>I'm looking for ways to 'tune things up', and have been thinking about
>this as an option. One thing that is a bit disturbing, the Admin Guide
>refers you to the Admin Reference for a list of server options, yet this
>option is nowhere to be found in the Reference. Wonder if that is
>significant?

>On to a more academic pursuit, in looking at the tapeiobufs option there
>is a note telling us that this option doesn't apply to 3590 device
>classes at 2.10 or higher (z/OS). My guess is this is because IBM has
>done away with BSAM (Basic Sequential Access Method) at these levels
>(for what I've no clue). Can anyone verify or shed any light on this,
>and offer any suggestions for performance enhancement? 4.1.4.0 seems to
>use a lot more cycles, especially for expiration and reclamation.

>--
>Jim Kirkman
>AIS - Systems
>UNC-Chapel Hill
>966-5884