I am not saying that this will solve your problem (I never tried id even though I keep it in my radar), but this project seems to implement something close to what Daniel is describing:
Some time ago I had to setup a replicated file system between multiple linux servers. I tried everything I could based on postgres, including large objects, but everything was significantly slower than a regular filesystem.
My conclussion: postgres is not suitable for storing large files efficiently.
Do you need that for replication, or just for file storage?
Alvaro Aguayo Jefe de Operaciones Open Comb Systems E.I.R.L.
all media files are stored in database with size varies from 1MB - 5GB
based on media file types and user-group we storing in different tables, but PostgreSQL store OID/Large-object in single table (pg_largeobject), 90% of database size is with table pg_largeobject