Hi, Le 08/04/2020 à 12:12, Donato Marrazzo a écrit : > Hi Laurenz, > thank you for your reply. > Are you aware of any performance drawback? We had a customer with millions of small Large Objects, in part because his application forgot to unlink them. As a consequence, pg_dump was using huge amounts of memory, making a backup impossible. It was with PG 9.5, I don't think the situation improved since. > Il giorno mer 8 apr 2020 alle ore 12:06 Laurenz Albe > <laurenz.albe@xxxxxxxxxxx <mailto:laurenz.albe@xxxxxxxxxxx>> ha scritto: ... > > I'm working on a use case were there are many tables with blobs > (on average not so large 32KB). > > I foresee that in 2-3 years time frame, the limit of overall blobs > will be breached: more than 2^32 blobs. > > - Is there a way to change the OID limit? > > - Should we switch to a bytea implementation? > > - Are there any drawback of bytea except the maximum space? > Don't use large objects. They are only useful if > 1) you have files larger than 1GB or > 2) you need to stream writes > > There are no such limitations if you use the "bytea" data type, and > it is much simpler to handle at the same time. +1 -- Christophe Courtois Consultant Dalibo https://dalibo.com/
