Microsoft Releases Computational Cluster Technical Preview To
Robert G. Brown
rgb at phy.duke.edu
Fri Feb 16 06:48:10 PST 2001
On Fri, 16 Feb 2001, Omri Schwarz wrote:
> On Tue, 13 Feb 2001, Mattson, Timothy G wrote:
> > Richard,
> > Please try not to be so negative. I was involved with producing the
> > "microsoft clustering solution", and I assure you, its much more than you
> > suggested.
> So far the only advantage mentioned
> is the faster C compilers available
> in the MS cluster.
> Cycles are cheap. Heartbeats are steep.
There is also the pure marginal cost tradeoff. Let's say that Microsoft
has seen the light and sells node licenses for ONLY $100. Cheap (but
powerful!) nodes cost anywhere from $500-$1000 these days, presuming
that Microsoft's nodes can run headless, which they probably can't but
we'll again be generous.
In this most favorable scenario, a cluster purchaser can afford one more
node in somewhere between 5 and 10, which is a 10-20% speed advantage on
code that scales well.
A more reasonable cost-per-node scenario would have MS clusters cost one
node in 2-3. No way the compiler advantage is great enough to be worth
Robert G. Brown http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb at phy.duke.edu
More information about the Beowulf