A Spark DataFrame looks like a table and behaves like a plan. It spreads across machines, waits until you need an answer, and then runs the work in one optimized pass.