The models are copied to the project tmp_cache when loaded (if not there already) and hard linked from there when saving setups. So your 10 iterations share the same copy and do not use 10 time the disk space.
4 Likes
That’s good to know, thanks @PhilippeJ
It sounds like that’s only within a single batchgroup so if you were using the same model across 10 batchgroups then the 10x still applies, no?
In the same project, all batch groups will hard link to the copy in the tmp_cache folder.
2 Likes
@PhilippeJ - what is the procedure for burn where a server may render a single frame at a time?
Does the temporary file get copied and deleted for every frame rendered?
A copy of the file will be in the setup that is packaged with the burn job that is sent to the manager and then to the servers. The servers will unpackage the setup that Burn will read, since Burn only load setups and does not write them, Burn will look in the project folder of the host and see that the file is there and do nothing.
Burn does not reload the batch setup/timeline in between tasks and keep it loaded until the job is completed or get assigned a new job.
In the case of background reactor, since the rendering is local the setup is not packaged in the job, so no copying done there.
1 Like
How about when doing a Flame archive?
When doing an archive you will get one copy per archive session inside the archive. The hard link should also be restored when restoring the archive.
For my earlier comments about the setups, note that if the file system used by user saved setups is different from the project setup file system, copies will be done each time since it is impossible to hard link between file system.