Doug's method works for the current design; but I would rather not have to
manage disk pools.  I would prefer a new feature that would allow grouping
nodes into a "collocation group".  All nodes in a "collocation group" would
be written to the same tape.  A node which is not a member of a "collocation
group" would be written to its own set of tapes (as its done today).
Naturally, multiple "collocation groups" could be defined.  Any thoughts?

On Mon, 24 Aug 1998   Doug Thorneycroft <dthorneycroft AT lacsd DOT org> wrote:

> One alternative to co-location. . .

> We are using several non co-located storage pools and one co-located
> pool. We limit the number of nodes in each non co-located pool to
> reach a good balance between The number of tape mounts required for
> migration, and the number of mounts required for the restore of a
> single node.
> Only our largest and busiest nodes are co-located.

[... cut ...]