A column-store will be magnitudes faster at analytical queries than any rowstore system. This is fundamental architecture and the data used makes little to no difference. You could use the exact ScyllaDB dataset duplicated to trillions of rows and still arrive at the same relative performance figures.