I have been working with SheepIt farm rendering for a while and see others commenting on some impediments that an enhancement to blender could help solve. I’m not sure of a better place to post this as I don’t know if there is a product management team to submit enhancement requests.
If one creates long particle or cloth or fluid simulations, there are folder collections of many small cache files, one per frame.
A farm often has a file size limit, in this use case the blender file plus the simulation cache folders enclosed in a zip file that could easily exceed this limit. And a farm often sends one or a small subset of frames to each helper (client) host. So a rendering project goes through “gyrations” of dividing up the farm rendering project into subsets of frame groups to overcome this restriction. Not a simple task, although scripting does help. But this effort detracts from farm rending just when a farm is ideal (large rendering tasks).
It could be useful if I were able to include a read-only, read-back cache reference to these that used URLs (ftp: or http:, etc). If the primary (read/write) cache exists (my local project machine), that would be used during render. If the primary (read/write) cache does not exist, but the URL reference was valid (i.e., I uploaded the cache to a publicly accessible server), that cache would be used. Call this the “alternative cache”. So when I submit such a vast job to a farm, all rendering hosts would end up accessing the alternative cache on my publicly accessible server (perhaps even on my Dropbox/Public folder), for the few small cache files they need for the frame they were assigned to render.