Last week,
Prof. Hal Varian visited BU to talk about his recent research using Google as the source for his data. While I did not take any notes :-), I found a
website with some notes from another talk by Prof. Varian that was very close to his talk at BU. One thing that I realized very early in his talk was that Google had a very sophisticated (theory driven) auction process for their advertising engine. Prof. Varian showed how this process was very robust and scalable. An interesting aside.. to a question from me about APIs, Prof. Varian did not think that Google would charge for API calls. However, he did leave the door open for charging once the volume of transactions from any source picked up. I am reminded of Eric Schmidt's concept of URL -- ubiquity first, revenues later!
Another presentation that I viewed (thanks to Prof. Venkat) over the weekend was a recorded
presentation by Paul Strassman about Google's system architecture. A key point that Strassman makes is that history may not remember Google as a search engine company, but as a company that created the architecture for large scale distributed data and application management leading to the realtime enterprise. To that end, Google has invested in hardware clusters, operating systems, and metadata management that give them high scores on reliability and response time.BTW, Strassman told the audience that programmers should learn API-related skills on core platforms to be relevant.