[racket] racket, scheme, and message-passing clusters?

Noel Welsh <noelwelsh at gmail.com> wrote:

> > My inspiration came after being able to quickly use Python's
> > "multiprocessing" library to easily take advantage of my multicore
> > workstation at work
> 
> You can take advantage of multicore already with futures. ZeroMQ would
> be valuable for distributed computing (multiple machines). Also check
> out places; I'm not sure where they fit yet (I haven't read the docs).
> 
> HTH,
> N.

Hi Noel,

I did see the entry for futures in the Racket documentation, but it
looks like it's targeted for a different set of tasks than what I'm
envisioning.  

For example, I'm processing large numbers of documents and would like to
be able to easily farm out some of the work -- for example, XML XPath
evaluations provided by libxml/libxslt -- to all of my machine's cores.
With the Python multiprocessing module, I was able to speed up execution
of some scripts by 3x or so.

That kind of task seems to lend itself to an explicitly shared-nothing
approach, rather than something like futures, which allow sharing if and
only if it is "safe."

I hadn't heard of places -- I only keep up with the released versions of
Racket.  The documentation at
<http://pre.racket-lang.org/docs/html/reference/places.html> seems as
though it's much closer than futures to what I want.

Thanks,
Steve