You should start with the Cluster config you have and how you managed to run these jobs with that config. Calculation and stuff and in what cases you felt that the memory should be increased with the real examples you came across.
You have not considered cores too.

This would definitely help the readers!!