The question is almost as meaningless as "how big is big?"
I suspect that the Social Security Administration of the United States has a database of every social security number ever allocated, and to whom, and name and all addresses over time and all employers over time, and all soc sec payments made by those employers over time, and current status, and so on. And therefore of every company that has ever had an employee, and who all their employees were every year and when they started and what their salary was each year, and what soc sec taxes were paid out on their behalf each year. And of course every social security check ever cut, and its serial number, and to whom it was sent, and address and date, and whether or not it was cashed or deposited or reported lost or stolen.
That one probably isn't very small, although it doesn't contain many 'blobs'. A 'blob' binary large object can be an image such as a JPEG, or a soundfile such as an MP3, or a movie such as an MPEG. It can even be another computer file or set of files and therefore it can be another entire database contained within a single field of a single record of the current database. I do not know if Sony Records has a database containing the track name, album name, performer, date, serial number, and the actual audio track itself, but they very well might. And for all I know, Paramount has one of every one of their film titles, properly digitized with scripted routines to auto-burn that content to DVD at any time or stream it for "Pay Per View".
It isn't farfetched to posit databases of several petabytes.
