Preferred cluster language...

[Robert G. Brown]

On Fri, 14 May 1999, Alvin Starr wrote:

> Is there not a perl module for PVM? Now there is a weird one. Perl for
> distributed programming.

Perl is also pretty easy to program with raw sockets.  On the one hand,
it's relatively slow compared to compiled code (pretty fast still, but
RELATIVELY slow).  The network, however, remains pretty much the same
speed.  This actually favors parallelizing perl applications, since it
makes the effective granularity of a parallelizable subtask bigger (more
time spent in the perl code block compared to the network IPC time).
However, one also has to face the fact that perl itself is not a sane
choice for jobs that will take hours or days to run unless there are
OTHER factors (like ease of programming, a need to handle lots of
regular expressions, portability, whatever) that make it so.

I have totally fallen in love with perl.  I like eclectic, kitchen sink
languages with human readable syntax and I >>REALLY<< like being able to
write conditionals and string morphs with inline regular expressions.
However, I'd say that perl is more suitable for "distributed information
processing" (the way it might be used to coordinate and filter a
distributed database on a distributed webserver) rather than
"distributed computing" (implying a compute intensive task).  PVM is not
a bad choice to support this, as it is likely to be some sort of
master-slave or task-pool sort of parallelism where speed is less
important than intelligence, and PVM is good for those paradigms.

Perl/Tk can also certainly perform lots of useful functions on the front
or back end of a "real" parallel program.  Perl scripts are great ways
to distribute and reap truly coarse grained parallel tasks, and perltk
makes writing a graphical front end very easy.  Anytime programmer time
is "expensive", perl is great because one can hack out and debug a
solution to many very complex problems extremely rapidly.  I'm hoping
that it is the last programming language I have to learn in this
lifetime, although I'm not optimistic.

    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@phy.duke.edu