clustering: how to start?

[Marc. Perisa]

On Wed, 01 Dec 2004 14:51:08 +0100, Devon H. O'Dell
<dodell at sitetronics.com> wrote:
> Koza Zbigniew wrote:
> [snip]
> 
> > Then I recalled that several years ago I experimented with FreeBSD and I
> > liked it. Now I hear that FreeBSD supports SMP "natively", so
> > no kernel recompilation will be necessary.
> > I thought - maybe we could try and use FreeBSD instead of Debian?
> > Now my question is: where should I start?
> > Which docs should I read?
> > Which program(s) should I use to cluster our 4 FreeBSD boxes?
> >
> 
> FreeBSD won't do process distribution natively. Want a project? :)
> 
> Other than that, there are other programs you can use to distribute
> applications in userspace (such as any of the MPI packages; I've had
> good experience with lam-mpi). To monitor the status of your clusters,
> you can use an application called Ganglia.
>

Hi,

the "normal" way to go is to use MPI for number crunching operations
on "normal" hardware. Since MPI is quite industrial standard most
programs which are spanning multiple machines support this interface.
To get a high real-world performance you should use high speed
interconnects with a low latency.

In general:
Are you writing the software yourself? Have you ever developed for MPI?
If not, then start reading at http://www-unix.mcs.anl.gov/mpi/

A general tutorial to setup such a system under FreeBSD may be found here:
http://www.bsdhound.com/newsread.php?newsid=205

Hope that helps

Marc