Hi Tom,
I think Eric's approach is the way to do it, i.e., dividing the
overall problem into units of computations of "PVs" and distributing
those units of work onto the grid, and probably it would be worth
re-using the serialisation code at least.
The couple of small items I would add is:
* When I was last involved in a project similar to this, we used a
commercial grid software vendor and I didn't think it was worth it
in the end, I would stick to an open source system like Sun's.
* You probably want to have the machine distributes the work sitting
on Amazon's network too, i.e., use one of Amazon's instances for
that. Therefore you will need two layers of communications.
Best,
Bojan
--
Bojan Nikolic ||
http://www.bnikolic.co.uk------------------------------------------------------------------------------
_______________________________________________
QuantLib-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-dev