Here is an interesting talk from PyCon 2008 by Peter Skomoroch (Juice Analytics).  Has anyone tried EC2?  In our local discussions it was said that EC2 is relatively expensive compared to other options, any comments?  I'm working on a smallish problem that would benefit from having a couple hundred processors for interactive analysis.  From the talk this can be done with ipython1 and ElasticWulf.  Here is the talk and abstract:

"Amazon EC2 may offer the possibility of high performance computing to programmers on a budget. Instead of building and maintaining a permanent Beowulf cluster, we can launch a cluster on-demand using Python and EC2. This talk will cover the basics involved in getting your own cluster running using Python, demonstrate how to run some large parallel computations using Python MPI wrappers, and show some initial results on cluster performance.

Slides available at

    4           5