I was wondering if you guys know whether there is a way to tell After Effects to use the global cache (which can be defined on a per user basis in the preferences) when rendering through aerender.
In conjunction with the “enable local rendering” I was hoping to get a better performance on the network.
We usually render a minimum of 10-20x version of our shots until they are final and the caching would optimize the rendering quite a bit.
I assume this is currently not possible, just want to make sure before requesting this @adobe.
As far as we are aware, this isn’t something aerender currently supports. These are all the flags currently available, which as far as possible, we already support in Deadline: helpx.adobe.com/after-effects/u … ering.html
On a side note, this page is super handy in tips to try and make AE go as fast as possible on your artist workstations: helpx.adobe.com/after-effects/u … rage1.html
Usual advice, like buy a SSD for local disk caching to be as fast as possible.
Would be cool if aerender did support this, as after the first hit, subsequent jobs referencing the same layers/footage should in theory be lots faster! If it could utilise any local GPU power as well…
exactly my thinking and one of the reasons i requested the “enable local rendering” for AE in the first place. (I also requested this fur Nuke just recently, let´s hope you guys find time to implement it more sooner than later).
All our workstation have caching SSDs and we put the global cache to good use.
We are currently working in 8k (horizontally) in 50p in AE and it really makes the server buckle.
The local rendering option definitely helps a lot but as you said, storing elements/assets in the cache would probably halfen the needed throughput from the server.
Let´s see what Adobe says (not much I am afraid, too much of a niche request).
I’m pretty sure if you have a render node with a Nuke local cache already populated with cache data, then when you network render in Nuke, it will take advantage of this cache, as it’s essentially still running Nuke, just with no GUI. If you ran a batch script which just opened up one at a time, all your nuke comp scripts on a machine and pressed ‘play’, the cache would get generated on each render node.
The issue with supporting local rendering with any of these 2D I/O heavy apps, is apart from the heavy hit you take the first time they run on a ‘fresh’ machine (let’s assume people are accepting of this hit, for the subsequent job speed-ups), is actually having to deal with data redirection & pathing. First you have the heavy File I/O copy of all the read footage, read geo, etc down to the local machine and also the automatic handling of file clashes which you are going to get and finally, do you render the write nodes locally and then copy back or the write nodes are left un-touched and render directly back to the file server, as we assume the heavy I/O bottleneck is in the ‘read’ performance.
AE project files are binary, so we can’t provide Path Mapping and re-direction directly, so you would have to export to it’s XML file format first. Nuke is easier here. I have ideas about how this might be possible, but it’s not a trivial problem
