I'm running on nodes that all have >1 processor. In the conf file for hadoop it says to set the number of reduce tasks per machine to a prime number close to the number of nodes, but should I instead be running close to the number of processors (or cores if I have 2 cores per processor)?
