I would interpret that as a compromise where C, A and P are divided into some ratio (you still don't get 100% of each). I understand Cassandra does that well if

I would interpret that as a compromise where C, A and P are divided into some ratio (you still don't get 100% of each). I understand Cassandra does that well if the application fits the model.