[Beowulf] SiCortex experience anyone?

[Joe Landman]

Jim Lux wrote:

> The value of a supercomputer under the desk is primarily to the end user 
> (faster whatever to free up an expensive engineer's time). They don't 
> see much difference between buying the software for a single processor 
> or multiprocessor.  The ISV won't get many incremental sales from having 
> a fast box to run it on, because the customer would probably buy it in 

Hmmm....  splitting proverbial hairs here ...

If the total cost of the solution is X(hw)+Y(sw) (hardware and 
software), and the solution cost on an "accelerated" box is x(hw)+Y(sw), 
where x << X, then the ISV may be able to convince the customer to buy 
at least incrementally more licenses of the software (which at the 
engineering shops we have dealt with, tends to be one of the major 
limiting cases for use).  Especially if N=Int(X/x) is large enough ( N>3 
or so), then the ISV can convince the customer to save money buy more 
licenses, on less hardware for about the same performance.  ISVs really 
want the end users to buy less hardware.  Or, more accurately, less 
expensive hardware.  This frees up more capital for ISV product 
purchases.  We used to see this happening at SGI all the time, when a 
"partner" would badly underspecify the hardware, just to make sure the 
customer had enough money to pay full price for the software.  Then we 
would bear the brunt of "its too slow" and wonder why they were told to 
buy an Indy or an O2 for their calculations ...

> any case.  The end user isn't interested in paying the ISV to port it to 
> the "supercomputer in a box" at least not on any scale comparable to the 
> actual cost of doing the port (i.e. they'd probably be willing to spend 

In most cases this is true.  In some cases, where there is a critical 
business dependency, this lack of port could have a non-positive 
cascading effect.

> <2x for the ported version, but they're also already forking out the 
> bucks for the box, too.. They tend to view that as the extent of what 
> they're willing to invest.

We live in the era of "good enough".

> 
> Jim


-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://www.scalableinformatics.com
        http://jackrabbit.scalableinformatics.com
phone: +1 734 786 8423
fax  : +1 866 888 3112
cell : +1 734 612 4615