Having an IBM mainframe would be great provided you have a direct
connect to the nearest power station! Here's several messages from
knowledgeable folks that posted to alt.folklore.computers awhile back:
[Message 1 of 4 in thread 148 of 176] alt.folklore.computers
Re: IBM Mainframes: Power & Coolant Requirements
From: Alan Greenberg <ALAN_at_VM1.McGill.CA>
Date: Wed, 23 Aug 95 20:33:11 EDT (Page 1 of 2)
In article <BNIDrJa.jrbrady_at_delphi.com>
Jason Brady <jrbrady_at_delphi.com> writes:
>
>Does anyone have any insights as to the power requirements for the
>beasts? How about the liquid coolant (I think one of the many system
>components was the PDU "power/coolant distribution unit"?)
We just replaced a two processor 3090 with a new CMOS-based system.
The new systems was cost-justified partly on the energy savings.
The processor complex ate up about 55 KVA, mostly in 415 Hz power
(which is what the motor-generator created from 60 Hz - just as
with airborne equipment, the higher frequency makes for smaller
DC power supplies). Onto that, you need to add the cost of
removing the heat - perhaps another 25-50% for a total power
consumption of about 60-70 KVA. A complex with more processors
would of course take more power.
I should point out that this was a 9 year old machine....
We replaced it with a new box that draw less than 1,500 watts!
Alan Greenberg
Re: IBM Mainframes: Power & Coolant Requirements
From: Joe Morris <jcmorris_at_mwunix.mitre.org>
Date: 24 Aug 1995 14:42:30 GMT (Page 1 of 2)
Jason Brady <jrbrady_at_delphi.com> writes:
>I never got involved with the facilities support side of the place, but
>remember one day when an emergency shutdown was attributed to a "motor-
>generator failure."
A lot of the big systems used MG sets, both as frequency translators (some
circuits were designed for 400 Hz supply) and as an isolation mechanism
to prevent garbage on the power feed from getting into the computer. The
rotating mass of the MG provided excellent ride-through for short outages
of the commercial power.
At a PPOE in the mid-1970s we had an IBM 370/148 (with MG), an IBM 360/65
(no MG) and a DEC DECsystem KL10 (no MG). We also had a Dranitz power line
monitor tied to the power mains; based on its output we could show how
well the systems handled dropouts in the power.
The DECsystem 10 never survived more than about 3 or 4 cycles outage; the
360/65 died at about 16 cycles, and the 148 could tolerate almost a
full second. (We got lots of opportunities to collect data points ...
thanks, TVA.)
Of course, the peripherals on the 148 would usually die on an outage of
one second, but the CPU itself would survive without taking a power check.
(The shop didn't have the funds and/or will to buy a UPS, even though I
managed to include a room for one in the building when it was designed.)
The problem with the DEC box didn't surprise me at all. There were a
lot of nice features in the design, but the impression I got was that
DEC designed the power circuits on a day after the engineers had spent
the previous night in a bar. At one time I found that the box was
dumping almost 8A of current down the *ground* line, yet the DEC
CE insisted that the box was installed and working correctly. It
also was dumping over 20A down the neutral line on a 3-phase circuit;
not the safety issue of ground current, but still indicative of poor
design.
Joe Morris / MITRE
[Message 3 of 4 in thread 148 of 176] alt.folklore.computers
Re: IBM Mainframes: Power & Coolant Requirements
From: Joe Morris <jcmorris_at_mwunix.mitre.org>
Date: 24 Aug 1995 15:04:21 GMT (Page 1 of 2)
Responding to: <41e5j3$id_at_news.iadfw.net>
mcw_at_gateway.airmail.net (Mike Wommack) writes:
>I used to work as an Electrician at an IBM site a few years back. I know
>the 3380(?) DASD strings took 220v at 100A. The plugs were huge. I also
>recall what looked like 3/4" water lines used to cool the main processor.
>(Which was either a 3090 or ES9000.)
If it was a US site, the 3380 box probably was being fed with 208VAC,
3-phase power at no more than 60A; the box is spec'ed at requiring
no more than 8.2 KVA. In the case of the 3380 the power connector
is an R&S 7328 plug, which is rated at only 60A.
(You could get the box wired for other voltages, so a 220 VAC feed
is possible but if that was the voltage the current draw would have
been even less than at 208.)
The size of the connectors in the power lines is often mandated by NEC
(National Electric Code) and/or local code requirements, which for
underfloor installation often require watertight connections (and thus
increase the size of the connector).
Joe Morris / MITRE
Received on Fri Jan 16 1998 - 21:11:45 GMT