I have some small hive jobs, extremely small. I.e. they are only taking 2 seconds or so to run, but there is no parrellization, only one node is running them. My available maptasks never changes, and only my CLDB node actually runs them. Is there a setting for small fast jobs that keeps them on one node or something? Is it because of how fast they are being spun up? I understand that I should try to make my jobs more efficient, but this is an edge case.... just curious what the issue may be. This is on mapr 2.0 with hive 0.9.