• One characteristic of "big data" I've noted seems to be that the dataset isn't easily contained and managed in a single server instance. "Sharding" comes to mind, and once you start splitting datasets into multiple servers, normal query techniques can go out the window. This isn't along the lines of connected SQL servers such that one server sends transactions to another, in my mind its more that a single relation simply cannot be handled by a single server, ie., a table lives on many servers.

    If all the data can be contained on one server, is it really big data, or is it rather just a VLDB?