Google for nginx, Cassandra, memcache, node.js, RabbitMQ, and visit highscalability.com. I'm sure folks will chime in with some other useful nouns and urls.

Highscalability.com has links to videos and articles of several major sites and their architecture.

The youtube video was quite informative.

On to the schema...without knowing your specifics, I would say to map your entities in a normalized form, but, you will most likely bust most of them out and denormalize as influenced by your display page (view) or particular portions of it.

Also play around with Amazon AWS. You can play around for several dollars an hour without having to invest in your own hardware.

Also, lurk around stackoverflow.com