Parallel programming

[jing]

On Wed, Aug 5, 2009 at 12:55 PM, William O'Higgins
Witteman<william.ohiggins-H217xnMUJC0sA/PxXw9srA at public.gmane.org> wrote:
> Yes - basically we have an agent-based simulation with an adjustable
> number of agents (but ultimately we will want to be able to run it with
> millions) and many, many scenario parameters - hundreds.  Each scenario
> can be run separately - no message passing.  However, right now a single
> scenario is being run on a single machine, and it is CPU bound and
> currently non-parallel.
>
> Finding available grids, clusters is not the problem - writing the code
> to utilize it is.

Just to add an additional plug for Snowflock... if you have lots of
existing legacy code and associated libraries and environments,
Snowflock works great (assuming Linux environment).  Because Snowflock
forks VMs, you just have to set up one VM image with the right
environment, add some minor tweaks to your code, and Snowflock does
the rest.  You don't have to pre-load the image on the other cluster
machines.  Snowflock dynamically streams VM images to other physical
machines to run.  I believe they currently have bindings for C and
Python.
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists