David Gibbs wrote:
The cheap variant: Roll back if it doesn't work within half the allotted
upgrade timeframe.
Sorry, Lukas, you really can't speak with authority on this ... you have
no idea what our situation was. There was no cheap or simple variant.
Once we started the upgrade we were committed.
I was involved with such an upgrade -- from a B70 to a model 300.
First complexity -- it was within a state government. Budget,
schedule, etc., were not solidly under IT control. They weren't even
solidly under the control of the same state 'Department'. Department
of Administration was the ultimate IT Executive Branch authority;
the Legislature was in 'control'.
Second complexity -- the model 300 would have a 3995 optical jukebox
to help manage the large number of imaging *DLOs. Brand new
technology added onto a brand new platform (for our group).
Third complexity -- the B70 had a 9337 disk array that was to be
migrated to the model 300.
Fourth complexity -- DASD on the model 300 would be approx one-half
the capacity of the B70. The *DLOs migrating to optical allowed
leaving _all_ of the old 9332s and 9336s behind. The model 300 would
have its base DASD plus the 9337 disk array (once the array was
migrated).
Fifth complexity -- the B70 was running near 90% DASD usage. This
was one of the driving motivations for upgrading to the model 300.
Note that a lot of the B70 single-level-storage necessarily included
the 9337s which would be gone from the system image soon. Going from
90% usage to a system with half the capacity _after_ a major chunk
of hardware is also migrated is no picnic.
All total, the model 300 didn't actually have enough base DASD to
accommodate all of the *DLOs and the 9337s couldn't be disconnected
from the B70 until the *DLOs were gone from the B70. The
save/restore cycles were done under _extreme_ care.
In short, once things started, there was essentially zero chance of
taking a backwards step. There was no "if it isn't working when you
reach half-way, then revert to the old" point. And it definitely was
not because IT was lazy nor because there was insufficient planning
by them.
Tom Liotta
midrange.com
