We're currently in the same situation you're in. Since we
don't have a SAN, my predecessor used the only shared storage he
had at the time: the database. Rather then upload all the files to
a folder, any uploaded document is placed in the database as a
BLOB. When a document is requested, a process checks if the
document exists on the file system. If not, then it retrieves it
from the database and places a copy on the file system. This way
subsequent requests from the same server return the file from disk
and not the database.
Thankfully, we're upgrading our infrastructure and our new
setup will have a SAN (plus CF8).
In the meantime, on Linux I'd have used
rsync to keep the
folder contents consistent between servers. After a bit of
Googling, I've found one
Windows program
that does the same thing. Might be worth a look.