The documentation says:
With MapR-DB tables, the cluster handles sequential keys and table splits to keep potential hotspots moving across nodes, decreasing the intensity and performance impact of hot spots. However, hotspotting can still hamper database performance.
But none of the courses nor the documentation makes it clear how the potential hotspot is moved across nodes. To the best of my understanding, tablets are stored in containers, so to move the hotspot across nodes would it have to migrate the tablet to a different container? That does not sound very efficient.
Can anyone shed some light on that?