I wanted to study how modern sites (Facebook, Twitter, digg.com, Flickr, etc...) scaled their architecture to serve millions of page requests. What was their initial infrastructure, when and how did they expand, and what motivated/justified their choices and solutions.
If you search on the web, there are scattered blog posts here and there, but is there a book or paper or article that documents some of the best solutions and case studies we've seen recently?
The High Scalability web site maintains an excellent archive of this kind of information:
http://highscalability.com/blog/category/example
for example, about the Facebook messaging system