How about writing the population to the file and using it as input to
map-reduce directly?  Evaluations that fit into a map can obviously handle
that.  Evaluations that need to be parallelized, can have the map emit
multiple copies with sequential keys.  This will put a copy of each input
into multiple reduces that can do a piece of the evaluation.  The pieces can
be put back together in a second MR step.

I even think that you should be able to demonstrate that any MR program that
evaluates a single population member can be applied to multiple population
members with slight modifications.

On Wed, May 28, 2008 at 11:44 AM, deneche abdelhakim <[EMAIL PROTECTED]>
wrote:

>
>
> I don't know if there are many ways to do it in Hadoop, but how about
> writing the population into a file and pass it with the DistributedCache ?
>
>

-- 
ted

Reply via email to