Briefing Materials?

[Robert G. Brown]

On Fri, 28 Jun 2002, Andrew Fant wrote:

> Does anyone either have any materials prepared for high-level briefings on
> what "beowulf" technology is and isn't or any pointers to places where it
> might exist?  I have to prepare something for an audience that
> instinctively things of all things VAX when I use the word cluster, and it
> would be nice to see what kind of illustrations and explanations people
> are using before I start reinventing the wheel.

   http://www.phy.duke.edu/brahma/beowulf_online_book/

is an online guide.  There are also two or three introductions to
beowulfery in the form of talk panels on the brahma website that you are
welcome to snag -- consider them GPL if you like (although attribution
is always welcome to academic humans:-).  I can provide you with the
latex sources if you want to use them as a base for your own briefing.

Let's see:  On the brahma site there are also links to many other sites
where useful material can be found; the main bewoulf page, for example
(www.beowulf.org), the beowulf underground, and many cluster websites.

Then there is the classic "How to Build a Beowulf" by Sterling, Salmon,
Becker and Savarese (MIT Press Scientific and Engineering Computation
Series), the beowulf HOWTO and the beowulf FAQ (all a tiny bit dated but
still relevant).  The general idea of cluster computing in general and
the beowulf in particular hasn't changed much since the early to mid
90's, although of course the underlying technology has moved from Suns
and Decstations and Irises on thinwire (or even thickwire, for those who
remember the good old Vampyre Taps:-) through linux, 10BT and Pentia,
into linux, multiple networks, and multiple architectures.  Put a pile
of COTS CPUs running linux on a common COTS network (everything as
fast/big/cool as you can afford, tuned CBA-wise to your project goals),
install PVM, MPI, or whatever you might need in the way of network-based
parallel programming support (sometimes "nothing" beyond standard linux
and tcp/ip) and do your work in parallel.

HTH

  rgb

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