Thanks for the input!

Adam's suggestion (using an UDF to read files) is an interesting one.
However, after moving a small set of images (about 140000) into a
database (and reading a few back) I find the "database approach" to
be very convenient. Currently the files are stored in a lot of
folders (some folder structures are quite deep!) because of having
been through a sequence of slightly different "solutions". Having
one, easy way to access the images (in the database) was nice. I
could make a new folder structure which might make the UDF solution
(almost) as nice, though.
Once scanned, the images are never changed. Some may be deleted, and
some may be replaced. Some may have their key data changed - this seems
easier with a database solution. (This - and the reasons why - will be
difficult to explain, so don't ask!).
There were no comments on the suitability (or not) of 16KB pages (as
opposed to 8KB/4KB).

Looking at the files again, the actual numbers are approx. 5.6 million
images (and 320GB). Testing the loading process shows that slightly
more than 20 images per second was copied from a fileserver to the
database. Fb/1.5.4, pagesize=16KB, Windows servers. The program was
run from a client pc, so the images probably did a fair lot of
travelling on the LAN...
As for backups/copies, there are several options in addition to the
obvious (but time consuming) one.


--
Aage J.