I’m designing asset loading for a project that will need to load many specific combinations of assets. So the approach of dividing assets into only a few .pak files doesn’t appeal to me. If I have a chunk with 20 trees in it and I know I only need the one tree, the bandwidth and space is wasted on downloading 19 trees I won’t use.
Ideally, I’d like to assign every single Actor or top-level Static Mesh to its own chunk, even if that results in thousands of chunks. I can then walk a dependency tree with the Asset Registry, and know exactly which chunks must be present on the system before this specific level can be opened.
I want to understand the drawbacks of having many chunks, in terms of processing and space overhead. Does anyone have specific experience with using a lot of chunks? Did you hit any walls? Did it become slow to traverse the Asset Registry? Did the manifest files become uncomfortably large?
I’d want to simply make an editor utility that periodically goes through the directory I’ve designated for streaming assets, and then assign a new, unused chunk ID to any asset that doesn’t have a chunk ID, or is a duplicate. I’d have to run this script proactively from time to time. Forgetting to run it simply means that unintended assets show up in the same chunk, or are distributed in chunk 0. But I’m open to other ways of automatically assigning chunk IDs.