[Bf-committers] Questions about cloud rendering

[Andrey Chursin]

Hi guys,


I am trying to create a cloud render farm for Blender (because there
is never enough render power!) and I have a few questions on how
Blender works. I have a solid background in distributed and cloud
computing, but I am still new to 3D rendering and Blender
specifically.


The few questions that I have are below. If it’s too much for an
email, perhaps there is someone who is willing to get on
skype/hangouts with me to help with these? I would be extremely
grateful! (happy to pay for your time too). Thanks a lot!


Questions:

1) I saw 'do not run scripts' in blender options. Does it mean the
rendering process would need to run some arbitrary code injected into
the blender file?

Are scripts commonly used? If we do not support it on the farm would
it turn off a lot of people?


2) I want to check files uploaded to my renderer before putting them
into queue. I figured I can do it by running:

blender -b filename.blender -P myScr.py

Can I assume, that if myScr.py was run by blender, that the file is ok
(more or less)?

Generally, what is the best way to check most of file's parameters
before launching computation-heavy rendering?


3) Does the packed blender file contain everything needed to render?

What’s the best way for the user to upload data to the farm? A packed
blend file? Or is it better to let them compress everything into one
archive?


4) Are all render settings already stored in the file? Is there
something that can’t be included in the file, that I’d need to ask
about before starting the render?

I see that I can specify output size, pick frames and so on from the
command line - but aren’t they already stored in the blender file?


Network rendering


5) I noticed (using wireshark) that the master starts an http server
and accepts POSTs with json commands from clients. These commands
contain local (to client) paths to files. So how does the client send
these files to master? Seems like no other data is transferred with
HTTP. Did I miss something?


6) I really like the idea of the network renderer, but right now it
uses network discovery and doesn’t have any authentication, so I can’t
really use it for my cloud farm. It seems like minor improvements can
be done in order to create a standard protocol and allow to push user
tasks to any farm (assuming the farm provides the user with some URL
like https://my.farm.io/ and username and password).

It seems like the only other solution currently if each farm writes
their own blender plugin.

Is there someone I can talk about network renderer? Perhaps, I can
even contribute.


-- 
Regards,
Andrey