I've been looking around the Beowulf sites for several days now. If you
follow ALL the links, you will eventually find the useful info. Basically,
Beowulf is used to enable several Linux boxes to behave as one
super-computer. That's the good news. The bad news is, this is only useful
if you are running programmes specially written to run on parallel
processors. How the system would behave if you just kept opening several,
sequential programmes, I'm not sure if it would automatically farm the
different programmes to the different nodes or not. You can, as noted in
earlier E-Mails, manually spawn these out yourself, but that is a bit like
buying ten copies of the same CD to get a 10% discount. You sound like
someone who just wants a plug-and-pray type system, so most of this mould
be out of your league. If, on the other hand, you would like to try to
programme your own concurrent code to utilise this Beowulf system, the
follow the links from Beowulf to the other super-computer sites and you can
find a wealth of info on building the system, programming the code and
build your own little super-computer. At one site I saw a Beowulf system of
16, '486 PC's networked together that had achieved 1.3 GFOPs! That's 1.3
Billion Floating-point Operations per Second! Quite reasonable from PC's
that can be bought for less than £150.00 @. This is just about the limit of
my knowledge on the subject, but I am planning to start piecing together my
own Beowulf system soon.
Cheers,
John Gay
