Re: They have doubled the number of servers and havereplacedthevirtual database with a high-capacity physical one

It sounds NOTHING like single level store to me. Don't get me wrong,
single level store is great but let's not try to be blinded as to other
capabilities.

Here's an example, I have a Domino server. It connects to DB2 for data
requests. So, it may do a 'connect to myibmi' or some such process. Now,
'myibmi' may be running on serverA or serverB, depending on whom Mimix
currently switched over to. Now, let's take this to the next level.
Assume both machines are currently up and running. Now assume instead of
dumb Round Robin or DNS spraying you have an intelligent device in between
that could sense things like "even though tcp/ip is running to serverA the
database is locked out for some critical processing or some such thing". I
use something called an Edge server for WAS fault tolerance for this exact
thing for WAS applications. Back to our scenario, it could then route to
serverB.
Take this to another level, we have both these machines. Why not spread
the load between the two? This spreader could route to both. Again my
edge server does this with my sametime chat clients between two sametime
community servers. DB2 on a Mimix environment is a little trickier. While
you may be able to query data on the 'backup' partition you cannot update.
So, you could use that as your query server (with some other
considerations like what if you only do daily backups from the backup
server and bring it down to do so?).
A day doesn't go by when I'll drop one or the other of my sametime servers
for backup or something and the users keep chugging along.

My IBM Edge server is pretty intelligent. Not only does it check to see
if port such and such is active on that server but you can set it up so
that when it checks a web page that a certain text is returned in the
page. I can't go into too much more detail as I've had a consultant set
up much of this.

DB2 Symmetric Multiprocessing is an option you can put on IBM i. It will
spread out database partitions (you may know them as members) between
different servers. Let's say you have a Region column in your table. And
you have an IBM i for each region. You can create a table in which it
will automatically know that if I write a row with a value of "East" into
Region then it will store it on the right IBM i. This may help regional
queries and spread the disk around (which isn't as much of a concern as
when it was in the days when machines could only support a limited amount
of disk). But it's still not to the level as multiple servers served by
an Edge server.

Rob Berendt