Normalize until it hurts, denormalize until it works. Build your database tier using everything they taught you in the classroom. Normalization is your friend

Normalize until it hurts, denormalize until it works. Build your database tier using everything they taught you in the classroom. Normalization is your friend here. Why? When you've effectively normalized your database, you can monitor the results of queries and eventually introduce the optimizations that you need to make things run faster. The problems that sites like highscalability.com deal with are far beyond the scale of most endeavors. If you run into massive scale problems, be very happy because it means you're building something that people want and you can afford to throw some money at the problem ;) I know people who have production databases with 30+ terabytes that is sitting in a relational database. If you design well, you shouldn't have any problems reaching those goals.