The paper that explains how Microsoft SQL learned to recover without undoing everything first
Under the Hood: The Research Behind Microsoft SQL, part 2 Here’s a scenario that used to be genuinely frightening: you kick off a bulk load of a few hundred million rows in a single transaction. Somewhere in the middle, the server goes down. Not because of your load, just an unlucky reboot. When SQL Server comes back, it has to undo everythin...