>        Now I have heard other things.  I heard someone on AIX saw that one 
> node
>would write to one volume and then the next node would write to a new volume
>but one node wouldn't split its data between two until it fills one.  Now
>I'm not saying this is true its just what I heard.
>        I also heard that in some platforms it just fills one volume and then 
> goes
>onto the next.  Maybe Richard or one of the IBM engineers could shed a
>little light on this?

Well, I am certainly not in the class of the IBM engineers: it's just that
I've been around a while and have experienced a variety of things.
I try to share what I've learned, keeping in mind that things can change
without notice and render old notions obsolete.

Writing distributed among storage pool volumes should be expected for the
direct access volumes (disks) in a single storage pool.  I would expect to
see one node per volume in the case of Collocation by Node, and Collocation
applicable to tape rather than disk.  There is also Migration, which tends
to operate on one node's data at a time.  But beyond that, I would generally
expect to see writing distributed over the disks in a storage pool.

Lik you, Kyle, I'd also like to learn of any platforms where this is not
the case.  I should think it would have to be due to some peculiarity in the
operating system.  Conversely, one would not expect TSM servers to be
developed for platforms which were not fully versatile.

  Richard Sims, BU