Henderson, TL Todd Todd_Henderson at
Wed Jul 24 12:21:39 PDT 2002

To me, there should be a single cpu version and then a multi-cpu version and
then license them based on number of jobs that can be running at one time.

We were looking at getting a 128 cpu cluster, roughly $1000/cpu.  But one of
the software vendors we were thinking about wants to charge an extra
12k/year to go from 64 cpus to 128 cpus.  That comes out to $562/cpu for the
3 year lease of the hardware for the extra cpus.  Thus, for us, it became
more cost effective to get faster, more expensive cpus and spend less on
software.  If you charge on a per cpu basis, then there is a balance between
getting faster hardware and getting more licenses of the software.  If
software starts to dominate the cost of a cluster, then I expect more money
will be spent on hardware.

Now, if however, it was 12k for the increase from 64 to 128, and then just a
few k, maybe 2-3k for annual licensing, that would have been different.


> -----Original Message-----
> From: Jim Phillips [mailto:jim at]
> Sent: Wednesday, July 24, 2002 12:39 PM
> To: Jim Fraser
> Cc: beowulf at
> On Wed, 24 Jul 2002, 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.
> First, would more processors make a single calculation faster, or just
> enable more 4-cpu calculations at the same time?  If it's more
> calculations then you would apply your standard volume discounts.
> If a single calculation actually runs faster, then it's a 
> question of what
> value that has for your customers.  For derivative pricing 
> there could be
> quite a premium, but none for calculating life insurance rates.
> -Jim
> _______________________________________________
> Beowulf mailing list, Beowulf at
> To change your subscription (digest mode or unsubscribe) 
> visit

More information about the Beowulf mailing list