IBM working on Watson app for smartphones

To answer your first question: Watson doesn't just search, it tries to understand all the information. It runs several hundred different algorithms in parallel and then weighs each of them for relevance. That said, you could get away much less hardware. For example, during the game the entire database was held in 16 TB of RAM because disk accesses were considered too slow. It had to beat Jeopardy champions in 3 seconds or less.