I just stumbled upon a blogpost on the weblog of Julian Hyde. Julian is discussing shared-everything and shared-nothing architectures and the way Oracle is handling things in this area. I just started ringing bells in my head about Massive parallel processing and how Google is working things like this. It is a great article you should read when you are interested in developing database grids and such. It is not a tech deep article as such, however it is giving a good idea about all those things you might already know but which you do not use and think about every day.
Good starting point to get back into cluster architecture thinking. Because I am currently getting back into thinking about how you can run a datacenter as efficient as possible with as much computing power per system as possible and the same time cutting down hardware costs and save some power this is quite interesting stuff.