Is that silicone talent or musical?  It's pretty easy, these days you 
can find packages that have most of the work done for you already. 
 Typically the hard part is figuring out what application the cluster 
will have (mass numbercrunching/rendering/compiling and what-not).  Mass 
rendering doesn't need much bandwidth so you can just send the data to a 
bunch of machines and let them go crunching, compiling is mostly data 
reads, so you'll want as fast as possible connections between the 
machines.  Mass compiling might also be more tricky in that you will 
probably run into problems with complex makefiles (see gnu stuff). 
 Rendering is the easiest application of a Beowulf or other types of 
clusters, so if you're just doing this for fun, I would recommend this 
path.  If you want to write your own code for it, I would say just go 
buy a book on PVM/MPI and read a small howto on clusters, and you should 
be on your way.   Good luck and have fun.


Kevin Anderson wrote:

>I haven't played with it, but I'd like to.
>
>Does anyone here have any experience with Beowulf (or other) clusters?
>
>What were your experiences?  On a scale of difficulty, where would it rank?
>I'd only have 2 or 3 nodes, It's mostly for curiosity...
>
>For the scale of difficulty, assume 1 is about the level of setting up a
>DHCP server.
>Assume 10 is finding the talent in the POP band of the month...  (N*Sync,
>Backdoor Boys, Britany Speers, etc, etc...)
>
>Kev
>


Reply via email to