I only recently started using Maestro, so it could be that I'm using it incorrectly. I'm trying to run a parameter space study that involves several thousand relatively cheap runs, where each individual run is serial. I followed the LULESH example for the parallel setup. What's happening is that Maestro is launching a new job for every parameter combination. This ends up being wasteful and slow, since each run is serial, so most of the processors on the node go unused.

Is there a way to launch a singe job on a specified number of nodes that will process all the parameter combinations?

Thanks