[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [datacenter] peak capacity planning for power vs. cooling

I would suggest the following. Yes, everything you are saying has value
especially from a physics class point of view, especially if you view the
little room as a calorimeter.

1) Unless you are talking about several hundred copper-delivered phone
lines, or more, you are dealing with very little amperage over the telphone
lines. I wouldn't be surprised if they act as a heat sync more than a heat

2a) Yes, air moving is heated by the fans, but not in a greater proportion
than the amount of energy being put into the fan to move the blades in the
first place. Since the energy in most of the fan calculations is included in
the input power for the machines, we'll considered it covered.

 b) Likewise, the contribution to energy transfer via a 10db lambda (fiber)
is negligible, though my guess is that the electrical equipment required to
power that emitter and fill it with data to send is not.

3) As for where the energy is going if NOT to heat. Well, some if it is
being used for the purpose intended. The purpose of this room is to do some
computation or something, and not convert 120A into pure heat.  So moving
the bits around in the registers of the computers generates some heat, but
part of it will be used in the act of computation (i.e. not going to heat,
but not destroyed either). I think for a .86A draw on some of our servers,
we figure .68A was heat based on actual testing by our manufacturer. I could
be totally off on that. If 120A of input energy were converted entirely to
heat, I would heat my home with servers and get the computation power for
free and more efficient than my furnace.

4) There is a factor we are also ignoring. That the draw of energy from
solid state equipment is slightly variable based on ambient temperature. The
_higher_ the ambient the more current (and heat) is needed (generated). I
think this is because the electrical paths become higher resistance inside
of the silicon wafer, but again I could be wrong.

5) Other factors we are ignoring. Depending on the placement of the AC/heat
rejection equipment, these items could add heat to our calorimeter. The fact
that a 4ton (48,000 BTU/hr) AC system may lose some of that cooling capacity
in the installation process through long coolant line runs and losses in the
ducting (air pressure, static pressure, etc).


Since monkeys much less intelligent than most of the people (I am making a
wager here) install, design and repair these HVAC systems and we all seem to
have generally decent temperature control, it is important to remember to
not overanalyze yourself to death and go with certain rules of thumb. If you
are putting 12KVA in a 400 sqft room then you are trying to cool 12000/400 =
30 VA per square foot. Convert that to BTUs (~100-150 BTU/hr per sq ft) and
any high school vocational protege can tell you how much airflow you need
and how big a compressor you'll need. Add 10% and you are there.


> -----Original Message-----
> From: Bruce Robertson [mailto:[email protected]]
> Sent: Wednesday, January 15, 2003 7:36 PM
> To: [email protected]
> Cc: datacenter
> Subject: Re: [datacenter] peak capacity planning for power vs. cooling
> > 100% of the power you consume is not used to generate heat
> Where do you suppose it is going, if not to heat?  The energy has
> to be going
> somewhere.  I could quite likely be wrong, but my reasoning on
> this is that an
> almost immeasurable amount would be going out your fiber as light, and a
> similarly small amount is going out as electrical energy for
> whatever copper
> circuits are leaving the area.  In fact, your telco may be
> injecting more than
> you're emitting, if they're delivering anything over copper.  Even the air
> physically moved by the fans turns into heat that has to be
> removed from the
> environment.
> It's been a long time since Physics classes, though, so please
> don't hesitate
> to correct me.
> --
> Bruce Robertson, President/CEO
>  +1-775-348-7299
> Great Basin Internet Services, Inc.			fax: +1-775-348-9412
> http://www.greatbasin.net

To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]