Given the following cluster configurations with equivalent compute power (100GB RAM and 200 cores), which one will generate the least network traffic during a shuffle operation?

//IMG//