What is the limit? (was RE: multi-homing fixes)

You're not the first to tell that i'm talking nonsense; but i had both
experience with a water-cooled computer (not something i wish to repeat),
_and_ designing hardware for COs (heck, the first computer i laid my hands
on was using thyratrones (sp?) for triggers, and quite a few relays).

Well, I really don't want to play "Can You Top This?" (The name of a national
radio program when I was a child), but the first computers of my experience
were mechanical analog machines--all gears and cams and such.

I think the word is thyratrons.

Not everything which works in one-of-a-kind supercomputers is workable in
your average central office. People who build equipment which has to work
for years w/o interruptions, survive quakes, power faliures and
thick-fingered technicans are not idiots; and if water cooling was a real
solution for CO environmental control woes, they'd use it.

UNIVAC 1100/90's and IBM 3090 K's (I am not sure of the IBM nomenclature,
I never actually worked on one) were hardly one-of-a-kind--there were three
or four 90's and at least a couple of the IBM machines in Hayward (look up
that place name in a fault catalogue) and Concord. The Bell System
probably had 50 or more /90's at the peak (how many 4A crossbar's did they

In an average CO a poodle on a floor has a risk of not being noticed for
a couple of months. It doesn't have an alarm relay, you see?

That would be pretty silly--I have not worked in a CO environment for
a lot of years, but I can't imagine one without such alarms.

(Have you ever seen a blast of H2/O2 mix resulting from electrolysis? No?)

All of the water cooled computers of my experience have the water inside
plastic plumbing. Unless somebody leaves the plug out of the bottom
of a tank.

And by the way, the CO's around here seem to get flooded some--tornadoes
rip the roof off, the Missouri disagrees with the Corps of Engineers, or
a bunch of elm tree leaves plug up a street drain.

The point here is not that liquid cooling electronics is the best, cheapest,
or even very good. The point is, the technology exists, and ut works.

That it is more expensive than air-cooling is pretty obvious. But I think
the facts are that the reason large-scale electronics are not liquid-cooled
is that in the late 1980's there were breakthrough changes in electronics
designs that greatly reduce the amount of heat produced per some useful
unit of measure.

If things continue as they are and absent another breakthrough change I think
we will be back to where, in order to do what we want to do, we will again
have to use liquid-cooling. And it will be an unfortunate expense, but it
will work, and it will allow us to do what we want to do.

My last on the subject.

What about using diamond to pull the heat away from these components instead of
water? Diamond is 4-5 times more thermally conductive than copper, can be put
into heat sinks in various configurations, and can give similar or better
results than water cooling when applied as a thermal transfer agent.

The trick to using diamond is managing the thermal junction between the diamond
and the additional material used to transfer the heat out of the box (because
obviously one can't create diamond fins). Diamond in itself is not a good heat
sink, but it is an excellent thermal transfer material. Point cooling can
easily be accomplished, and this has already a proven solution for laser diodes
and the like.

IIRC diamond is already being used in some Cisco products.

I suddenly understand! Every cisco memory chip is decorated with
diamonds, which is why they cost $7000, rather than $50!


All you need is a charcoal briquette and some nanotech.. :wink:

OH wait... nanotech is in the same category as COTS PC hardware that will
handle OC-192 I/O loads on its poor little PCI bus... :wink: