With all due apologies to David Lethe, who threatened to sign off if people
continued to tell all these "non-serious" war stories, my favorite story
about an HP3000 follows.
Bruce Toback writes:
>Marise Gwin writes:
>>battery backup kicked in. Twenty minutes later we're up and cooking - no
>>VREPAIR, no reloading of NLMs, no re-mirroring. Just go. ...
>
>>I love telling those stories to the non-3000 world. Too much, I'm sure.
>> Too smug, too much of a UNIX -basher.
>Again, _standards_. The user's data is precious; let's treat
>it as such...
>It is AMAZING what users of other systems put up with.
>-- Bruce
>PS. Admittedly, that was in the days when all HP-supplied
>software fit on one 1200-foot 1600bpi tape!
Our first HP3000 was a Series 33 purchased in 1977. It's still here in my
office and theoretically capable of being turned back on, although it's now
used as a desk and its 7925 disc drive is now being used as a printer stand.
About 12 years ago, when the machine was about 5 years old, all of a sudden,
the modem closet began filling with dense blue smoke and the smell of burning
resistors filled the room and eventually the building.
Quite obviously, something catastrophic had failed. But it wasn't easy to
determine what. We ran a time-sharing service at the time, with about 25
remote customers all whanging away on the 33 (with its massive 1/2 meg of
memory). Although the room was now quite smoky, everything was still working.
The computer was still up and working normally. And as best we could tell,
every modem was alive and well. And the phone system was working.
We couldn't tell what the heck just burned up. After an hour or so of
looking, we gave up; we figured we'd find out sooner or later.
That night, as we always did at 7:00PM, we performed a full backup. And just
a few minutes after we finished, the Series 33 very politely shut itself down
with a power failure.
What had happened was that a number of components in the power supply in the
Series 33 had basically evaporated, but in the 33, the power supply didn't
directly supply the electronics cabinet (CPU, memory, ADCCs); rather, it only
charged the batteries. So when the power supply blew up, the machine went
merrily along for the next seven hours using only its internal Ni-Cad battery
power.
What made the finding the problem all the more difficult was the speed of the
fans that cooled the power supply. They exhausted the air from the power
supply so fast that we couldn't see the smoke coming out the HP3000 -- and
all that smoke simply welled up in the modem closet, which was nearby.
We are all electrical engineers here, and always have been, and having almost
all of the components in-house, we disassembled the HP3000 to get at the
power supply, rebuilt it (redesigning it a bit to fit the components we had
on hand), and re-installed it. And, of course, the Series 33 came right back
up where it was, without a hitch.
I was a fan of the HP3000 long before that incident happened, and it only
further solidified my love of this machine. For business purposes -- where
data integrity is everything -- there is no other comparable machine or
operating system that I would ever recommend to a business customer. We have
databases we created in 1977 still running today -- and in 17 years of use,
we have never lost one bit of data or suffered any form of database
corruption. That's a tough record to beat.
A good portion of that integrity clearly lies in the province of software
(operating system, database), but no small part lies also in the design of
the hardware and its integration into the operating system and database. All
systems are going to fail sooner or later; the Second Law of Thermodynamics
guarantees it. The real question is how is the system going to handle
critical business data when the machine does fail? The old HP3000s have set
the standards high enough that I believe that HP may even have trouble living
up its long-term users' expectations.
Wirt Atmar
|
