[Bf-cycles] [GSoC2014] Introduction and some questions regarding an idea concerning cycles.
tim at criteriondigital.net
Mon Mar 17 22:07:40 CET 2014
My question is, does the current distributed render feature built
into blender not already support cycles ??
If you are looking for a centralized management for a true render
farm, then I would recommend warewulf. It is an open source cluster
suite for Linux of which I am a contributing developer. It has
historically been and RPM based solution, but my latest contribution
was porting Debian support to it. The latest release contains this
support. Warewulf also supports stateless nodes, which require no hard
Warewulf is open source project from Berkley Lab's at University
There are numerous ways to implement frame processing across a farm.
I do not know how Blender's net render feature functions under the
hood. Whether you have a team of artists who spread render time across
they workstations or you use a dedicated cluster, I assume basic
net render feature should be adequate for general use.
With all the options already available, I don't see a need to reinvent
the wheel. I do not use the net render feature in blender, am using
custom code for specific resource management across the cluster here.
Maybe I am missing something ?
On 2014-03-16 15:00, Lukas Stockner wrote:
> just some idea I had recently: Currently, as far as I can see, the
> networking is done over RPC. Using MPI would probably be useful, since
> there are many well-tested impementations of it and it's basically the
> standard on clusters. Of course, porting the Network device to MPI
> would be quite some work, a MPIDevice would also be a possibility.
> Just my thoughts,
> Am 16.03.2014 20:03 schrieb Matthias Herster <hermac03 at gmail.com>:
>> first of all, a short introduction of myself:
>> My name is Matthias Herster (usual nick: macro23) and I'm a student
>> of computer science at the University of Technology Graz. I am
>> currently in the last semester of my bachelor's degree and want to
>> apply for GSoC-program at blender, because I've used it now and then
>> for some little projects and want to help to improve it even more.
>> After over ten years of coding experience in a lot of languages, but
>> especially C/C++, I think I'm up to the challenge and have a realistic
>> point of view for the effort of such projects.
>> Now to my questions, regarding the idea of a network distributed
>> cycles rendering, to refine my proposal/application:
>> + Although I looked through the code in the past days and have done a
>> possible bug fix (T37973), I think I could need a brief overview of
>> which functionality regarding the network rendering of cycles is
>> already implemented, just to be sure.
>> + What should be the final goal/functionality of this particular GSoC
>> idea in more detail? (or is it up to me to make detailed suggestions?)
>> I hope someone is kind to answer my questions.
>> Matthias Herster
> Bf-cycles mailing list
> Bf-cycles at blender.org
More information about the Bf-cycles