Parallelization with Python: which, where, how?
mmueller at dgfz.de
Tue Dec 21 14:04:36 CET 2004
> Can someone recommend a parallelization approach? Are there examples or
> documentation? Has someone got experience with stability and efficiency?
I am successfully using pyro http://pyro.sourceforge.net for my
parallelization task (8 node Linux cluster, embarrassing parallel).
It comes with a good documentation and it is easy to use. It works on
clusters as well as on heterogeneous networks. You just need to start
a server on each cluster node (e.g. rsh node1 startserver.py). You
also need a name server running (maybe on your master node). Then a
client can just asks a server to calculate and gets some result back.
You can also send an object that calculates on the server and comes
back including the result.
There are some examples included in the documentation you can start
with and just change the parts that are specific to your application.
It is easy to understand and use and has been working well for me for
more than 2 years.
More information about the Python-list