Dual CPU nodes?

Robert G. Brown rgb at phy.duke.edu
Mon Oct 21 08:28:09 PDT 2002


On Sat, 19 Oct 2002, James Russell wrote:

> I have recently become interested in grid/distributed computing.  I
> want to construct a small (4 node?) cluster for learning purposes and to
> help with data mining for an M.S. class I am in.  I was wondering if I
> could use 2 dual processor (Tyan, maybe?) motherboards and get the same
> results as 4 single boxes, since my room is cramped.  Can anyone help me
> out?  Also, what kind of power requirements will I need?  Thanks!

Depends on the task.  For lots of tasks, dual processors are fine and
cost optimal.  For others, they aren't.  You can look at the book online
at www.phy.duke.edu/brahma to learn a bit about how to figure out your
own particular tasks.  For a learning cluster, duals are probably
acceptable anyway, although you'd find it easier to "recycle" single
CPUs to your own and friends desks after your class.

Most standard circuits have enough capacity to run two duals or four
singles - you can estimate that they'll draw somewhere in the vicinity
of 400-600 Watts, which is roughly a third the capacity of a twenty amp
circuit for a switching power supply load, more like half the capacity
of a 15 amp circuit.  Then you've got a monitor and network switch,
which can be another 100-200 Watts.  A beefy UPS is not a bad thing to
consider if you can afford it.  Finally, remember that you have to get
RID of all that heat.  You'll have the equivalent of a small space
heater running 24x7.  If your room is small, and you have 150 Watts of
light bulbs on, your own body (100W or so), your cluster (call it 750
Watts although it might be as low as 500 Watts) you could be dissipating
750-1000 W of power continuously (about the heat dumped in by one of
those small electric radiant heaters).  Your AC has to handle that --
you'll need a thermostat IN YOUR ROOM and it will need to WORK to remove
the heat or your room will stay around 80-90 F.

For that reason you might well want to select systems based on their low
power requirements rather than their speed...

   rgb

> 
> ---------------------------------
> Do you Yahoo!?
> Y! Web Hosting - Let the expert host your web site

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu






More information about the Beowulf mailing list