-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Well, you have the option to use the bytea types. It's nice to have the stuff
in the database. I wrote an application (quite a while back) where I had to
store documents to the database. At that time I switched to mysql, not
because it's better, but because it handles binary data quite easily. Since
this is a couple of years ago it's quite likely that PostgreSQL byte handling
improved a lot (back then PG was still at a pre 7 version).
Meanwhile I'm developing an application that stores huge amounts of documents.
Since this time I need the features of a real database, like triggers etc.
I'm using PG now. The amount of data I'm storing now is huge, therefor I
decided to store the files in a directory tree. I'm storing about 40000
documents around 100k each a week, thats about 12GB a month. The file storage
directory is handled by the server side of the application and is invisible
to normal users. The main reasons for using the filesystem instead of the
database were:
1) a lot of documents
2) due to 1) I'm able to spread the documents over several
harddrives/filesystems - so this solves the storage space problem (unless you
want to use LVM or similar)
3) much easier handling on the application side. I can use rsync, ftp, scp or
whatever else to transfer files. Also most programing languages are quite
good at handling normal files.
4) You never run into a file problem regarding file type or size, at least not
up to the limit of the operating system you're on.
5) You can use conventional backup methods without interrupting the database
server. It's also much faster to restore the whole system on a different
machine - dump/restore the database and copy your directory tree.
You have some drawbacks though:
1) your application has to make sure that the referencing records and the file
storage is in sync. If there's a bug in the application you can totally
scramble the storage.
2) If your server has a filesystem problem you could lose files due to fsck,
which again would "unsync" the data from the files.
Basically it all depends on the structure of your application. If you can
guarantee that only the application has access to the file storage you can
control the synchronization problem. If not I'd probably go the extra mile to
store the stuff as blobs.
There is another option you might want to consider. If you're storing
relatively small files, like the images for a website, you could even go with
a normal "text" field and base64 encode the data before storing (as well as
decoding it after retrieval).
My $0.02
On Saturday 21 February 2004 10:53 am, Holger Marzen wrote:
Hi all,
AFAIK it is possible for columns to be very large, up to about 2 GB. Are
there any hints or experiences about storing binary data (jpg-images,
pdf-documents) in PostgrreSQL with or without the complicated lo-stuff?
Of course it's in many cases a good approach to store those files simply
in the file system but there's always a risk of running out of sync
(filesystem and tables), e.g. by deleting files and not deleting the
table rows with the filenames.
Any ideas and comments welcome.
---------------------------(end of broadcast)---------------------------
TIP 7: don't forget to increase your free space map settings
- --
UC
- --
Open Source Solutions 4U, LLC 2570 Fleetwood Drive
Phone: +1 650 872 2425 San Bruno, CA 94066
Cell: +1 650 302 2405 United States
Fax: +1 650 872 2417
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)
iD8DBQFAN8g4jqGXBvRToM4RAreNAJwMwT1Twtg9c35rp1Sgag hKU7XDiQCg0bQM
Z7cbgYmGYtjkCFnJfSL3tm8=
=3yhL
-----END PGP SIGNATURE-----
---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?
http://www.postgresql.org/docs/faqs/FAQ.html 
