Can anyone point me to some documentation as to how LiveJournal scales out to multiple back-end SQL servers? Is there some kind of data partitioning (e.g. by user)? If so, how does a feature like 'Friends' that seemingly needs to query the entire back end get implemented? If not, is there some kind of replication scheme (e.g. master and slave database servers?) How is load balancing amongst the database servers achieved? In this world, is there any plan for the point when LiveJournal gets big enough that it is physically impossible for all the data to be stored on one SQL server, or for all the write load to go through one SQL server?