Jim Meyer purp at
Wed Jul 24 10:55:29 PDT 2002


On Wed, 2002-07-24 at 11:49, Jim Fraser wrote:
> With regard to scientific software pricing for clusters, what do you guys
> consider reasonable model for pricing?  Lets say the basic price for some
> given software is 50K for 1 CPU. What do you think is reasonable for
> increasing numbers of CPUs.

I think you'll find that the value proposition for a given piece of
software is closely tied to where it offers the greatest cost savings.
It often seems to come down to whether or not the software ties more
closely to the value of the user or the hardware. 

A good example from my industry (digital animation) is a
modeling/animation package vs. a motion blur plug-in. The animation
package is very closely aligned to the value of its user; the better the
package, the more work they can get done. That's why an average price is
~$7K/seat with support and maintenance going for another $1K/seat

The motion blur plug-in is more closely aligned with the value of the
hardware. While it will save the user a little time and perhaps give
them a simpler way to get a particular look, they will only use it
interactively for a scattering of frames in their animation; the bulk of
its use will be on the render farm, where the economies of scale apply
and spending $10K more to save ten seconds per frame may not be as
generally worthwhile as purchasing five more $2K machines with which to
get the user's work done by morning dailies review.

Of course, this is a gross simplification. You can work it out
mathematically by playing with software costs, average salaries, machine
costs, MTBF, etc. to come up with how it applies to your particular


Jim Meyer, Geek At Large                              purp at

More information about the Beowulf mailing list