Ideas

Like the idea,

there is a huge potential market here.
Simplicity & compatibility will rule !
I would suggest to make it a webservice with a generic interface that can easily be called fromany tool.
Providing a VPN would be to complicated, have the output available on a server untill it is checked out.

I am not sure I get it completely. But i think I understand the basics. I vote yes.

Ian, what confusion do you have? Basically rendering is to animation source files what compiling is to source code... it's the process of converting the instructions into a final product which gets viewed. It's a very CPU-intensive task. Companies like Pixar can afford to have their own massive inhouse renderfarm composed of a ton of blade servers usually. For the average animation studio however, it's cost prohibitive to have an inhouse renderfarm. Since rendering is usually the bottleneck in the animation process, people outsource their jobs to renderfarms. RenderNow.com is one such outsourced option. I'm proposing to compete with companies like RenderNow using cloned virtual render appliances running on Amazons ECC service. The only development that would need to be done would be to add the billing/job tracking infrastructure and create an easy interface for submitting jobs right from the animation programs (ie. no going to a website and uploading source files & specifying parameters).

Does that clarify it?

Sean

I love this. Top rating from me. This is unique, innovative, POSSIBLE TO IMPLEMENT (not all that common among our idea cloud here), and would really be a valid service.

Rock on, scrollin'. I'm all up in this s#$@

well you go guy and make that idea work cause its sounds great and best of luck to you and let me know when you are about to make alot of money cause i could be you personal assisantance take care good luck

If Muster on ECC works, then this idea might have legs! I'd like to see a proof of concept!

I'm not a rendering expert. One place I worked they blew $14,000 when they made an error subitting their rendering job to a render farm. It's a commodity service - price sensitive. How much cheaper is it going to be than And you're sandwiched - you have hard costs with amazon EC2 and if your customer doesn't pay you, aren't you stuck with the bill? Could you target selling the entire business to amazon as an add-on/service wrapper to their EC2? They would be in a better position to eat costs from bad credit. How much cheaper is it than RenderNow? It could be first-mover sensitive.

Why not combine this with distributed computing, and sell off the CPU cycles of folks who sign up - and then pay back for those cycles?

I've been waiting on a good, profitable, use of my spare cycles for years...

@vulgrin- yes, combining this with distributed computing would be cool but that's not core to the project- it's a way to optimize once things are working and making money. Right now that would be an added layer of complexity given we would need to handle the more-important challenge of making the product work and evangelizing it.

@Dug -
>they blew $14,000...
that sounds like either pilot error or poor customer service on the part of the render company but not a flaw with the model.

>It's a commodity service...
exactly. but the utility computing aspect gives the ability to disrupt the incumbents that have have to manage huge banks of servers, update systems, pay big power bills, pay admins, etc.

>How much cheaper is it going to be...
this would need to be researched. My guess is that much of the cost of the existing services goes to overhead that we would not have.

>you have hard costs with amazon EC2 and if your customer doesn't pay you, aren't you stuck with the bill?

no, you do not have fixed costs with EC2- it's pay-per-usage so those are all marginal and can be rolled into what you charge the customer. as far as customers not paying: set the billing up with pre-payment - load your account up and draw down as you use it. funds are captured before the final renderings are released. This is how RenderNow does it.

> They would be in a better position to eat costs from bad credit.

this issue goes away with the above suggestion.

>How much cheaper is it than RenderNow? > It could be first-mover sensitive.
needs to be researched. Not sure what you mean by "first-mover sensitive" - there's certainly room to partner with someone who's already figured out the job submission piece - this would be an excellent candidate except that it's windows-based software-> http://software.fran...x.aspx?page=deadline

Find the equivalent on linux and that's a no-brainer.

Solid feedback. Keep banging on it - this is a good dialogue.

-sean

I really want to see this one go the distance. Go dubs!

This is a great idea!

Combining it with distributed computing would be great, though potentially fraught with licencing issues. (Everyone participating needs a copy of the renderering software, which is expensive, so your audience becomes limited - unless it's cost-effective for people to buy a licence just to participate! - and how do you know they haven't pirated it? If they have, does that make you complicit in the crime?)

Chris,
thanks. actually the licensing issue you describe wouldn't be a problem if we were to leverage something like xgrid on the mac platform as people would not be running the render software locally, just donating CPU cycles... Again, i just see the distributed computing aspect as a wholly-separate project... make a platform that allows people to monetize their CPU cycles then ride this and other CH projects ontop of that. wait quick somebody submit that one! ;-)

Ironically enough, the JumpBox technology we've developed would make a perfect delivery mechanism for deploying the render VM's into Amazon's EC2 service. I'll see what it would take to throw together a Muster JumpBox and do a proof-of-concept on the Amazon service.

Sean

Gets my vote.
Both ideas are likely to work, but this one is very much in the spirit of CH.

I should point out that others have attempted to sell cpu cycles. See this article from 2000:

http://www.geek.com/...gee2000630001780.htm

Both companies are folded up it seems.

Here is a recent project that I saw on Digg:

http://www.cpushare.com/technical

Yep; a service like Kinko's to render animation would be very widely used.

thanks Mark, motiggidy and karmagoddess-

I've been in contact with both Amazon Web Services as well as Vertex (makers of the Muster render software) and have thrown together a page on my blog that gives a brief Executive Summary of how I see the Renderfarm opportunity for anyone who wishes to explore it in more depth-> http://www.scrollino...om/ideas/renderfarm/

Hopefully this clarifies things a bit as it seems some people have confused the renderfarm concept with distributed computing. In-depth questions or feedback can be directed as comments on that page.

exciting stuff... thanks for everyone that voted and happy holidays.
sean

Sounds great, can't wait to see how I will be able to get onbard with this one to make some money for the ol retirement fund that is non-existant, lol. Hey if we build this then maybe we can make that other guys movie as the next project using our custome platform, what a perfact way to test and get the name known a a major player... mmmm need to keep this in mind...

The idea is a good one, however I think technically will be very difficult to execute with EC2. Here is my quick take:

1. EC2 does what Muster does. It takes an AMI (an Amazon Machine Image) a virtual server and runs it transparently across the many physical servers that make up the cloud. So I fail to see how Muster would help. I would think it (Muster) would not have access to the physical layer below the virtual AMI to actually dispatch to individual cpus/clients the way it is programmed to do now.

2. Assuming then that EC2 can do the job that Muster does and transparently distribute the render jobs, licensing of software like Maya, 3ds Max and other render software/engines would be an issue. They either license per hostid/primary mac address or per cpu/node again none of which are accessible from the virtual environment only the underlying Amazon Grid OS (for lack of a better term) would have the information to the actual physical info that the licensing code would need. If they are not accessible then the licensing code will prohibit the rendering to work. And if it is accessible then would the cloud be licensed by the number of cpus ? (the render software guys would think so I'm sure) if so then the licensing costs would be a huge capital initial expense.

I would suggest talking to Amazon EC2 and asking about how this idea would work technically also talking to the software vendors. They may not have an idea on licensing since nothing like this exists yet but you would have to have them on board I would think to make it work.

The other issues has been mentioned but I think are actually easier to overcome - creating an interface for users to send jobs, tie to billing etc. That stuff is as difficult as making any other web app or interface which has been done before and therefore the knowledge exists.

Not saying it is not possible, just wanted to provoke more thought in the hope that this idea can be made viable and profitable.