The Renderfarm Project Poll

General Forums Blender and CG Discussions
1

Hello,
As some of you might have noticed I started a discussion a couple of days ago regarding starting up a Blender-specialized render-farm. My intention was to try an provide a render-service for the Blender and OpenSource community as cheap as possible, preferably free, but after having done a lot of research myself and gotten advice from several others on the forum I have gotten to the conclusion that a purely free render probably would be impossible, and if possible it would be very hard to run, it would be crowded by freebies and thereby be much slower than a service you payed for. So I have worked to merge this into a conclusion where you get some frames rendered for free but if you are rendering large amounts of frames such as an animation or something with a lot of different frames you would have to pay. I have also gotten other suggestions. Initially I do not want to run a home-computer-cluster based farm, what I wish for the whole community is a stable and well-run traditional farm in order to maintain and be able to control the quality of hardware and software ran on the farm. Some have also expressed a wish for not having to install clients or other software on their computer in order to be using the farm, a thing I agree with for now.
But as I have been discussing and heard a lot of peoples opinions I have been able to narrow down a few ways in which this could be executed, thus the poll.
Even tho i narrowed it down to the solutions mentioned in the poll I am still more than willing to learn and listen to advice from the community as you in the end hopefully are going to be the users of this farm. I ask you to be honest in your answers in the poll and if you have any issue at all I wish for you to reply to this thread so that I can look into it and we hopefully can come down to the best solution.

For those of you wanting more information, here is the original thread:
http://blenderartists.org/forum/showthread.php?t=184985

and here is a letter I have been circulating to potential sponsors:
http://uploader.polorix.net//files/248/OpenRenderfarm/letter%20to%20potential%20sponsors.pdf

If you have advice, questions, meanings or anything else you want to say about this project please tell me so.

Bjørn Ihler

2

For those of you wishing to pay, what do you think would be a fair price pr render-hour/18GBram and render-minute/18GBram or pr month?

(yes I am thinking of selling render-time in chunks of 24GB ram (using three nodes of 8GB with quad-core or processors with as many GHz as possible with the option of increasing and decreasing the number of nodes you use)

Bjørn

3

I have started communication with a few potential partners and sponsors, if you have suggestions for people or companies to contact please tell me so. Don’t let this project die before it has even started:)

4

i have gone for the last option, mainly because im a fan of open source, and am unemployed

but i could stretch to paying depending on the cost

5

Well, I believe most of us are fans of open source as we all after all are blender-users. The goal of this project was initially to create a free renderfarm and I haven’t yet ruled that out as a possibility, but it would be hard to run a good service on so limited funds. The point of this poll is to figure out what the end-users want, when I have more information from the potential sponsors etc. I will create a chart with price indications and indications on how payments will be done. I hope to hear from as many of you as possible in order to create the best service available for the blender users.

6

I voted for 2 as I’d only use such a service for renders that require uge amounts of RAM which I don’t do every day (or even weeks or months)

7

[ x] I have another idea on how to do this and have stated so in the discussion below

I like the idea of having a professional run blender dedicated renderfarm and would use it.
I also like the idea of cloudsourcing and sharing within a community.

Personally I would try to realize the renderfarm in the most flexible way possible.
I keep the explaination tabular for the easy reading:

  • Native and professional hosted machines for the renderfarm as a solid base.

  • Additional client machines in a cloud to contribute to the renderfarms base.

  • A free render option with low priority, maybe a framelimit of 10 mins and 1800 frames per 24h which equals 1min HD so BlenderBoi101 can render his bouncing cube with 5 particles within 1h and not wait 2 days on his celeron notebook.

  • Members that pay can raise their priority and framelimit in the renderfarm and got no framelimit per day, which can be split in several packages for purchase, also in monthly fee packages or 24h packages, or even a point debit system… buy 1000 Burnpoins use them whenever you need them.

  • There should be a clientsoftware where you can contribute FLOPS and become part of the cloud, but unlike in vSwarm, which in my personal opinion has a complete failsystem, you do not gain priority by contributing which degrades over time, but Burnpoints, which you can use lateron for your tasks.

The economic thinkers already see the problem though with that system. People will contribute, stash Burnpoints and when they need the power, they use their points to unleash the force of the full farm, while the operator of the farm sits in the dark because he can´t pay for the electricity anymore because he got no income.

So you need a smart system to make sure you have a fair give back for the community to contribute to the farm, maybe use the earned Burnpoints to purchase packages cheaper, or limit the usage of burnpoints to XXX frames per day.

However I think the key, especially for blender is a mix of contributive cloud and native farm. I guess I would use it. I got ~40 Cores to share and sometimes it is still not fast enough. If I want a draft to preview the animation I rather have 400 cores than 40. Thats where I´d use earned Burnpoints. Once sure it got production quality, I would most likely be willing to pay to get the job done “overnight” while the community can benefit from the 40 cores rendering their cubes =) I benefit for being able to render fast and when I need it, I would pay, or pay less for contributing.

You could rate the machines in the cloud and add the more potent once to the farm itself and use the less potent ones for the community cloud.

And I think it should also offer a variety of raytracers. blender internal, luxrender, yafaray and the usual suspects.
A downside making the farm now is that we got openCL/cuda on the march, so it might be intresting to make a GPU farm for octane/luxgpu…

What I also would like to see is a slider which impacts the price for rendering. Sometimes I don´t need full throttle. Something like a pulldown menu to choose:
Q9550, 8GiB is my system (your database holds the rating for such a machine)
one frame takes me: 12’50’’ to render
amount of frames: 9000
date and time due: tomorrow at 12:00

and then the farm offers a package of XX machines that approximately get the job done by then, but cheaper, as I don´t need the whole farm.

There problems can raise too. What if a guy buys a high priority job after I bought my tomorrow 12:00h job. You would not be able to finish the high priority job.

So basically I got lots of wishes and ideas which lead to lots of problems, which I believe is not the first time they show up. But as this is a discussion forum… discuss :slight_smile:

that was my bag of 2cents :slight_smile:

8

@arexma: As you say this is a discussion forum, so discussion is of course welcomed and I value your bag of 2cents greatly.

As of now I think this project will start at one point with a fairly small but efficient farm and then develop from there based on the users wishes and economical possibilities. Thus it can evolve into having a gpu-render-part, a cloud of clients et cetera is things which might and most likely will develop in time as a natural progression for the farm-service. As the service you describe might as you say end up with me in the dark it would not really benefit anyone. A renderfarm without power is about as useless as soup without a spoon.

The idea with sliders which will impact the render and the price sounds like a great idea and I will try to embed it in the web-page where you buy render-time from the start.

@all:
What I first of all need is money to get the farm started so if anyone have ideas for funding please tell me about them, I am e-mailing potential sponsors and supporters but if you know anyone who might be interested in sponsoring or supporting this in any way please tell me about it.

Bjørn

9

is this thread really that un-interesting?

10

Ok, so I sniffed arround a bit and got 3 nodes, each with two Quad-core processor(Xeon E5430) (with the posibility of getting a 4th in the future) for free, only thing I need is to get some(a lot of) RAM, so if you have suggestions and comments, please come with them=)

11

I would not mind paying a monthly fee and have the Render Farm Administrator balance the load out.
How does it get balanced?

12

The key to making this work, economically, is what Arexma already mentioned:

A free render option with low priority.

That way, you do a great service for people who are really hard up, and,
for those willing and able to pay,
you give them the best possible inducement to actually pay!