As one with more than a passing familiarity with unix server design and sys-admin issues, I found this in-depth article describing google’s server architecture a fascinating read.
While the emphasis is on the hardware design choices, there is also quite a bit of info regarding the information flow through the system, from indexing the web to serving those results to customers. Their reliance on cheap, commodity hardware might be surprising in light of google’s overall reliabilty, but their info architecture is such that failures of individual servers are basically invisible to the public.