>Try prototyping applications using map-reduce for huge datasets.

Yep, a Spark job I was running recently used up my Mac's 16GB rather quickly.