[Bf-funboard] Re: Use idle online internet computers as a distributed rendering

[Ton Roosendaal]

Hi,

> Hi Ton,
>
> This is my last post on this subject.

Urm... what does that mean? I would really like to encourage people to  
pursue development of a good distributed rendering system. Whether or  
not it will be internet based is more of a technical detail... it can  
be included in the specs, but the initial emphasis (roadmap) can still  
be to make it work in a controlled local environment first. This is  
what our artists currently need most.

This mailing list is not really meant to discuss implementation or  
technical design. It also doesn't make a lot of sense to pin down such  
requirements for a distributed rendering system when nobody actually  
takes the responsibility to execute it.

If you are motivated to work on this project, you have all freedom to  
define your specifications yourself. Then use this mailing list to get  
feedback from Blender users about the functionality it offers.
For inclusion in our 'official' distribution, having support from the  
people here then is a basic condition, but it should not stop you from  
working on your own ideas. Sometimes presenting people a working  
solution really helps to get acknowledged. :)

Also!
There are at least 2 other projects here about this topic:
http://projects.blender.org/projects/networkrender/
http://projects.blender.org/projects/nitrox/
You might contact them, to find if they're interested to work with you.  
If you need technical feedback (or just want to hint developers) the  
information you sent below really belongs to our developers mailing  
list, bf-committers@blender.org.
http://www.blender.org/mailman/listinfo/bf-committers/

BTW: If you like to use our projects site, you only have to ask me! All  
projects related to 3D and/or Blender are welcome there.

Thanks,

-Ton-


>
> Obviously, you know your product and customers better
> than me.  Most of the concerns mentioned are bandwidth
> related.  I'd just like to make the following
> additional points:
>
> - Bandwidths are only going to increase and become
> cheaper going forward.
>
> - I would encourage you to implement the most general
> purpose architecture possible.  Putting in place an
> architecture that falls in place in the future is a
> good thing, I think.
>
> - I think BOINC could be used as an intranet solution,
> then opened up later as bandwidths increase.
>
> - It looks BOINC has done much of the work already,
> though it may be more complicated than you really need
> at the moment.
>
> There are three similar solutions that I've found that
> could be investigated.  They are:
>
> Berkeley Open Infrastructure for Network Computing
> (BOINC)  Open source project.  My guess as the most
> likely candidate.
> http://boinc.berkeley.edu
>
> United Devices - Grid.  Commercial product.
> http://www.ud.com/home.htm
>
> Entropia - DC Grid.  Commercial product.
> http://www.entropia.com
>
> I might be interested in helping with this project if
> one of the goals was high quality output and another
> was to somehow harness a large cheap source of
> computing power, like with BOINC.  I'm a C++
> programmer with over 10 years experience programming
> under various OSes, and I graduated as a Mechanical
> Engineer with honors.  I would help with the BOINC
> part, since working on the core of Blender is probably
> beyond my capabilities.  I think bandwidth may not be
> as much a problem here in the US, but that is only a
> guess.
>
> I'll probably have this e-mail address for quite
> awhile, but sometimes I change it, so I've included my
> snail mail address below.
>
> Sincerely,
>
> Robert S Townsend
> 8233 Verbena Drive
> Riverside, CA 92504
> USA
>
> <Hi Robert,
> <
> <We've been playing with such ideas in the past
> <several times, most obviously the problem is in the
> <difficulties of controlling such a public service:
> <
> <- animations will easily run into 100s of megabytes
> <of results.
> <Bandwidth limitations (uploading such amounts) will
> <make it difficult for most users to participate
> <- there is no method in Blender to detect or control
> <the amount of resources needed... someone can easily
> <make a .blend file that uses 1 Gig or more memory
> <- Blender files itself, packed, with all textures,
> <also can grow into 20 or more megs. Another
> <distribution problem
> <
> <But most important:
> <If you do an actual calculation; taking into account
> <the amount of time it needs to transfer/distribute
> <all data, render it, and get it back, you'll easily
> <and up with unacceptable delays for such a system to
> <work efficiently. Blender is really an animator's
> <tool, designed to quickly render huge amounts of
> <frames overnight. For example for a high-quality
> <renderer - like Yafray - this might be much more
> <efficient.
> <
> <My suggestion:
> <Let's first develop an easy & simple system with a
> <nice GUI for intranet purposes. With as main
> <features having a good control & overview of what's
> <going on with all rendering jobs. In practice,
> <Blender renders so fast that one artist can easily
> <get his work rendered overnight at 2-3 systems. Most
> <people have a spare machine hanging around. Or small
> <studios then can efficiently use the workstations to
> <act as renderfarm in spare hours.  With such a
> <system in place, it will be mich easier to review
> <possibilites to extend this to an internet service.
> <
> <-Ton-
>
>
>
>
>
> __________________________________
> Do you Yahoo!?
> Yahoo! SiteBuilder - Free, easy-to-use web site design software
> http://sitebuilder.yahoo.com
> _______________________________________________
> Bf-funboard mailing list
> Bf-funboard@blender.org
> http://www.blender.org/mailman/listinfo/bf-funboard
>
>
------------------------------------------------------------------------ 
--
Ton Roosendaal  Blender Foundation ton@blender.org  
http://www.blender.org