Out of core data sets (32 GB - 20 TBs). Data sets larger than 32 GBs cannot fit in main memory, as a result we need special algorithms to deal with data sets. Please aggregate data sets of this size here. I'm assuming commodity hardware, if anyone has access to a server then these data set sizes are completely irrelevant.
Data sets larger than 20 TB will probably need multiple computers and make use of message passing interfaces. If anyone find data of this size please post them here.