When I watch a big job run on my cluster, I expect the data being written to /tmp/mapr-hadoop to be more. Instead I see lots of reads and lots of writes to my mapr disks rather than my local disk. I NEVER see more data than say a couple of MB written to tmp, even running terasort. Something seems off here when if you consider MapR wants us to 10GB + free on on our /tmp. Is something not writing correctly locally? What settings control this? I know I am going on gut here, but something feels off.