Intel is finally shipping the 64-bit Itanium

Josip Loncaric josip at icase.edu
Tue Jun 5 10:36:31 EDT 2001


Jeff Layton wrote:
> 
> One ALWAYS needs to
> benchmarks their app(s) on all potential cluster solutions.
> I really don't care what I run on
> as long as I get the most speed for a fixed price.

Same here.  We estimate performance by looking at benchmarks that
resemble our code, then pick the best price/performance solution.  This
happy situation is typically found a generation behind the 'bleeding
edge'.

We stay close to the Moore's law by expanding/upgrading a section of our
heterogeneous cluster each year, with the expected lifetime of nodes
being about 3-4 years.  Annual upgrades fit our budget cycle and help us
track the optimal price/performance curve more closely.  Usually, we can
get better price/performance by buying replacement nodes than by
retrofitting old ones with upgraded components.

Itanium is a 'bleeding edge' product today.  It is not price/performance
competitive at its introduction, but in a couple of years prices should
come down and its 64-bit address space should start looking very
attractive...

Sincerely,
Josip

P.S.  Picking the best hardware replacement interval is an analytically
solvable problem.  Assuming that performance/$ doubles every N months,
the most cost effective policy is to buy replacements whenever you can
get 4.92155 times the performance for the same money.  The Moore's law
says that N=18, so the best replacement interval works out to be 3.44867
years.  Using intervals of 3-4 years is almost as good.

-- 
Dr. Josip Loncaric, Research Fellow               mailto:josip at icase.edu
ICASE, Mail Stop 132C           PGP key at http://www.icase.edu./~josip/
NASA Langley Research Center             mailto:j.loncaric at larc.nasa.gov
Hampton, VA 23681-2199, USA    Tel. +1 757 864-2192  Fax +1 757 864-6134

_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list