[Beowulf] RE: Capitalization Rates - How often should you replace a cluster? (resent - 1st sending wasn't posted ).

Nifty Tom Mitchell niftyompi at niftyegg.com
Thu Jan 22 19:53:07 PST 2009

On Thu, Jan 22, 2009 at 08:26:15AM -0500, Robert G. Brown wrote:
> On Wed, 21 Jan 2009, Lux, James P wrote:
>> I think there's a B-school thesis in this topic. It sure would be
>> interesting, especially if you had a decent modeling tool that would let you
>> plug in cost of money, funding profiles, etc.
> discussions.
> This is right in the sweet spot of model complexity.  The dimensionality
> of the problem is order of ten -- not unmanageably high -- and there is
> fairly solid data to support model parameters.

To begin, a more modest model might be network storage.
Disk capacity cost and power.... service contracts etc.

On the HPC CPU side some postings have been written that imply it is more
cost effective to postpone the purchase of a cluster to a point that Zeno
would be happy to play.  i.e. a cluster today might solve a problem set in
four years and a cluster that could be purchased in two years for less
money could solve the same task set in two years so doing nothing is
the most cost effective choice.   And yes, in two years doing nothing will
still be the most cost effective solution if .....  ;-)

This Zenoesque analysis ignores the value of the result which is mostly
unknown.   Further uncertainty and tax liability implies that it should
be carried on the books at zero dollars further validating the model.

	T o m   M i t c h e l l

More information about the Beowulf mailing list