Running test001, I noticed that barely anything was happening on either of the 2 nodes on this test system. My laptop fan isn't even running, and it always runs.

My initial guess from watching top is that there's a lot of chatter involving the controllernode. Since test001 is made up of mostly small queries, my guess is that there's a lot of back-and-forth getting session #s, getting unique #s, & the other non-versioning functionality that went into controllernode eventually, which is required to init a joblist.

We should look into why we have such low utilization when there is a constant stream of queries being run, and then make another ticket for proposed fixes. One possibility mentioned in Rome is that we're opening & closing connections too much. Maybe make the connection a singleton or a pool in the DBRM class. Maybe a singleton that becomes a pool when there is contention. Or magic. Magic always works right?