Just an aside note:
 
I know that many of our CEs (at least here in Germany, and
local procedures might vary) use(d) a utility called DTDUTIL,
to avoid INSTALLs at customer sites, in case the mech has
to be replaced with the controller but still contains all
valid data.
I do not know, if DTDUTIL works with the mentioned 2GB SCSI
discs, it copies the disc content to a tape/DDS (the latter
is awfully slow, AFAIR, but at the time I dealt with this stuff
there were no SCSI DDS drives avail), and after replacement
one could copy the content of this disc back 1:1.
Faster than an INSTALL on BIG volumesets, slower on small ones;
and this is how I would keep my system_volume_set in terms
of changing user data: very small.
 
That doesn't mean I would not support the idea of mirrored
system volumes for real high speed recovery, RAID technology
does not cover controller or host-adapter problems.
 
But what about doing some 'hand-mirroring' of your own?
Just have your second set of system volumes sit there,
hooked to a S918, mirror your changes daily per STORE/RESTORE
or more often via Network (if you really change files in that
domain that often), ready to be plugged in , if the 'big one'
fails (may be on an unused SCSI or FL card in your main system
to add that redundancy as well).
If one system volume fails, you will anyway have to reboot
(from a different config group).
 
That would require much smaller changes to MPE as the support
for mirroring system discs:
- the support for spoolfiles i.e. IN/OUT.HPSPOOL on user volumes.
  (but you could hand-mirror them as well).
- the ;DEFAULTVS option on the NEWACCT / ALTACCT command,
  since with above concept you really only want to keep the
  system accounts/software on the system volume set.
 
There is no need to mirror transient space, it is gone after the
necessary reboot anyway.
 
Sounds like a new product ? :-)
 
Goetz.