On Fri, 2002-11-15 at 19:10, Dion wrote:
> Fascinating stuff, have you got some kind of compute intensive 
> application to model the performance of the cluster??  Thats the part 
> that has me buggered, finding a high compute load task to put such a 
> cluster to the test.

I wanna see distributed factorization with a number field sieve. We
could break RSA properly and take over the world!

http://mathworld.wolfram.com/NumberFieldSieve.html

Trouble is there's about three implementations of it in existence as far
as I know, and the one I have is closed source (I think the others are
too). I'm guessing Mosix won't make serial code parallel, so we'd have
to have a code-fest first I guess.

But yeah, it's a great test. Factoring a number that's only a few
hundred digits long brings my Athlon XP to its knees for quite some
time.

There's lots of hardcore computer problems though - I know Sydney Uni
has a course devoted to just such problems. Travelling Salesman Problem,
and all of its buddies for example. Even boring, known worst case
algorithms are cool on clusters - bad ass image rendering for example.
The nice thing is that that's easy to split up for any number of
processors.

My 2c.

James.

-- 
SLUG - Sydney Linux User's Group - http://slug.org.au/
More Info: http://lists.slug.org.au/listinfo/slug

Reply via email to