How to run a memory intensive spark streaming job using the smallest memory usage possible

Question

I am running a job that joins two kafka topics into one topic using a value. The environment I am using only allow me to assign less than 10g ram per job, the data I am trying t