Stan Sieler wrote:

>With the 9x8 series, HP dropped the internal batteries in favor of
>an external UPS.  This was said to have been done because of the
>increased power requirements of the newer systems.  OTOH, I suspect
>it was to allow them to produce a cheaper machine for some sales
>situations ... sort of like a car manufacturer saying "you can save
>1% if you cut out the seatbelts".

I must respectfully disagree with Stan here (with his interpretation of
the facts, that is, not with the facts themselves). The disappearance of
memory-only battery backup seems a little disappointing, especially since
us old-timers can no longer do those fun demonstrations where we yank out
the CPU plug and the system doesn't miss a beat. But in fact, the
replacement of memory-only backup by whole-system backup is a significant
improvement.

When HP first started including battery backup for the memory subsystem,
in 1974, that was actually the best that could be done. The 512K memory
subsystem in the Series II took as much power as a whole system does
today, so the only economical powerfail protection scheme was to protect
the memory contents, and have software save CPU registers contents and
other context information when the power failed. When power was restored,
the power-on software would detect that memory was intact, reload the
context information, and continue processing. This elaborate scheme is
what was so much fun to demonstrate. If the power failure outlasted the
energy in the battery -- usually about 15 minutes -- the system would
fail and would have to be restarted manually when power was restored.

Today, a reasonably-sized battery can supply enough energy to keep a
whole computer system running for as long as the old battery backup
facility could do no more than maintain the contents of memory. So where
the old system would suspend processing while power was out, the new
system allows processing to continue during the power failure. This is
especially important in today's computing environment, where many
computers are used by people far away from the place where the power is
out.

Again, if the power failure outlasts the UPS battery energy, the system
fails. But that's no different than with the memory backup scheme.

I agree with Stan's implied point that having the UPS as an option may
lead some customers to omit what should be seen as a vital element in the
HP 3000's reliability features. On the other hand, the MPE/iX transaction
manager, which also didn't exist back in 1974, will prevent serious
damage to the user's data in the event of a power failure. There really
are some applications where hardware powerfail protection is unnecessary.

Further, the use of a UPS allows options that weren't possible with
memory-only backup. It should be possible, for example, to implement a
scheme where a user-specified command file gets executed when five
minutes of power is left, and another is executed when one minute of
power is left. (There should also be corresponding recovery files for
both these points, if the line power comes back on before battery power
fails.) Such a facility can completely eliminate the risk to user data
from an extended power failure -- something that couldn't be contemplated
at all with memory-only battery backup.

-- Bruce