Developing Rendering scheduler - need your help

Hi Blender Community,

I am a grad student in computer science at the University of Miami, and I am wondering if some you from the Blender community can help me with my current project by sending me some of your Blender projects.

I am working on a script that analyzes a Blender projects to predict how “long” they will take to render on given computer. I think this script will benefit our scheduler for our network renderer.

I am trying to gather a collection of Blender projects that I can use to test (and teach) our emerging heuristic. While I certainly can create my own projects to test the system, the problem is that I generally use similar rendering settings when I create my own Blender rendering, andI want it to be more general purpose and it should work on projects created by a variety of users. In other words, I trying to gather projects that reflect the rendering settings that Blender users have in the real rendering projects. And I think I can only get that by asking others for their projects.

So, if you have any projects that you could send my way, that would be great! ([email protected]) I am looking for projects that are all set to capture rendering times with my test/learning system. Your help will certainly make my program more complete (and I promise not to share your projects with anyone). Also, until 2.5 become stable release, we will be developing the system from 2.49 internal renderer.

At the end of the day, I hope to make the script public: a simple python script that could run within blender to get a sense of how certain parameters might effect the rendering time. But, the particular use will be for render farms schedulers (sending hard projects to faster computers and easier projects to slower computers). If the project goes well, it will of course become open sourced.

Please contact me with any questions, and thanks in advance.

Adam

[email protected]

Now I’m not a scripter- so I may not necessarily know what I’m talking about- But 2.4x is, as Andrew Price put it, sort of an “old hat”. Its API may not change, but I think that as soon as 2.5 is released, pretty much everyone will switch to that (I think the majority has switched already). I think that developing it for 2.5 could give you more of a “running start” and more testers/interest.

Just my 2¢.

dsavi, you might be right on here. I current network render that we use to test out different scheduling/distribution algorithms works with 2.49. I developed it that way because that is the stable release. But, it might be time to move away from that.

So, based on your recommendation, I would just say, please send me whatever you want: 2.49 or 2.5. And we will try to figure it out from there.

The idea is to create a learning system that observes variables in the rendering settings of a project, and then to be able to make some kind of prediction of how long it should take to render. Well, actually it will not give an actual time, but a type of hardness scale, such that a set of projects could be sorted from easy to hard based on what the system learns over time.

Now, to do this, we need to observe variables that people actually use in the real world in their own projects (it would be silly to observe possible setting, as some will emerge more prominent than others).

I do not have access to many other animators, that is why I am asking for assistance. I want to see what types of settings people actually use in their projects, and then create the system to observe those characteristics.

So, 2.49 or 2.5 projects would be great…

Thanks,

-Adam