I would say this solves this by allowing you to batch segments of that bulk... a streaming bulk transfer may not be possible because you don't know what server

I would say this solves this by allowing you to batch segments of that bulk... a streaming bulk transfer may not be possible because you don't know what server the data is on, and what order it comes in, etc... Google has had a lot of success building web applications using this generic, split-up way of storing data... it sucks for one box, but for their cluster of so many, that makes up for it. I haven't had problems with the example queries like what you're asking for, if the dataset that it modifies is small.