What is the main cost of a reduce-side join?

A. Both datasets may be shuffled across the network B. It cannot join large inputs C. It requires one mapper only D. It removes duplicate keys automatically