We recently set up an Xsan environment for one of our clients, and have run into some unexpected behavior with After Effects rendering. Without going into all of the dirty details (at least for now), one of the end users noticed that when he would render an AE composition to a local secondary drive, it would be faster than when rendering this same comp to the Xsan volume…his workstation is 3.06 GHz, 6-core Mac Pro, 64 GB of RAM, Mac OS 10.8.4, and After Effects CS6. We did some initial tweaking of the disk cache, memory, and multiprocessing preferences, but the same discrepancy in render times held true.
We then built out a test Mac Pro with just the OS and CS6 suite installed (we did not build out the client's workstations, so we wanted to eliminate other potential variables). We initially conducted both Finder copies and BlackMagic Speed Test benchmarks, both with read/writes to local drive around 134 MB/s and 700 MB/s to Xan (what we'd expect).
We then created a 30 sec comp in AE using source ProRes HQ footage (1920 x 1080, 24fps) with no effects applied; we rendered this comp out in three different formats (ProRes, Uncompressed, PSD Image Sequence); in all three tests, it took longer to render to the Xsan than the local drive (with the image sequence taking considerably longer):
ProRes = 1:18 (Media) vs. 1:24 Xsan
Uncompressed = 1:29 vs. 1:48
PSD Image Sequence = 2:13 vs. 3:41
As I mentioned above, tweaking preferences would improve overall render times, but those would still remain longer when the Xsan was the target destination.
Have other users noticed this? Are we missing something obvious here? It just seems unusual that an Xsan volume that can support 700 MB/s of throughput, does not show this performance when rendering from AE. We're also going to do some testing from Premiere Pro and Final Cut Pro 7, but since the client primarily works in AE, that's where we want to focus most of energy.
Any feedback would be much appreciated!!