I'm trying to learn to use Grid::Cluster to run some processes on our compute-cluster at work. We have 18 machines, 17 can run 8 processes concurrently and one can run 16 processes concurrently. But I'm not the only one who uses the cluster, so in my old, inefficient bash script to do this, I used to check how many processes other people were currently running by parsing the information in /proc/loadavg/, and then start only so many myself that their processes could continue undisturbed. How does Grid::Cluster determine when the max number of processes has been reached? Is it based only on the processes I start in the script that uses Grid::Cluster, or does it take the current load of the machine into account?

I'm also going to have to find out how to "nice" my processes. Maybe I can solve this on my own, but any pointers would be welcome...