Behind the Scenes at RebusFarm

An interview with Ralph Hutchermann, founder of RebusFarm

CGA: Please introduce yourself. What is your background and how did you get into running a render farm? What made you want to get into this business?

RH: It all started when I was studying Mechanical Engineering in the early 1990's. I got interested in the production of music video clips. I shared my ideas with some other enthusiasts among the German diaspora in a small university town, far away from all professional production environments. We soon realized that our desire to start professional media production would only be possible if we worked with 3D animation. We only had a very limited budget, so this solved all our problems with the heavy costs of lighting and video equipment. Also it made us independent of that nasty trap, amateur actors.

We were part of a new generation of film makers. In those days, almost all film production was done using analogue video equipment and/or expensive computer hardware. For instance, a Silicon Graphics Workstation with Alias Wavefront (the predecessor of Autodesk Maya) could cost €100.000 or more, not to mention the Betacam Recorder and so on. Instead, we began to produce our films using 3DS Max, with DOS and Video-Cut in Windows, working with conventional PC Hardware. 

Immediately after the end of my studies in 1994 I founded a small studio in Cologne, the capital of German media production. The studio worked freelance on numerous projects. As the company grew we had to expand our render farm continuously. But however many machines we bought, we never had enough. And yet, most of the time those machines just stood around switched off. What a waste of money! So we contacted some other studios and leased them the machines for a few bucks when we did not need them. 

What with one thing and another, the renting of the machines to other studios grew faster than our 3D Animation business, so we began to focus on offering a render farm service. We began with just 20 machines or so in 2006, which was of course, not really much, but since then we have been able to double the number of machines every year. 

I began very early to examine the process of Job submission from a local Computer to a render farm because I was not satisfied with the state-of-the-art. In those days it was usual to collect all of your Render job files manually, pack them and send them by email to a render farm. Some competitors developed web interfaces for manually sending the packed file to the render queue, which is clearly better than sending it to a supporter by email, but it was still not good enough in my opinion. I was thinking of a system were the 3D artist could send his render jobs to a farm in much the same way he does it with local jobs: Just press the render button!

I got the idea of developing a plugin for all 3D softwares that would collect all the resources of a job, would validate that the job is compatible with the render farm, and if not, would simply give error messages or better than that, perform tweaks to the job to make it compatible. The unbeatable advantage of the idea is that you will never send a job that renders locally at your desktop but fails at the render farm. Also we further developed the management software, which is not our invention but we did include, for example, a feature that saves the final renderings to a directory on your local computer. Our aim is to come close to an experience where the user cannot see any difference between a local and a remote rendering.

We have also published apps for Android, iPhone and Windowsphone so you have control of the render’s progress on the go, which is a step further to help our users to avoid unnecessary costs for failed renderings and lets them leave the office earlier, always a good thing!

Time passed and we solved complex issues in the background of our render service and our system improved. Meanwhile a lot of our competitors adopted our concept, but without getting their software to the same level.

New entrepreneurs in the cloud market, such as Spotify or Dropbox, provided new User-Interfaces to link users to their cloud services. I was inspired by their approach and added some of my own ideas to the mix. I divided the manager into a web-based ControlCenter and RebusDrop. The ControlCenter lets you manage your jobs from any device and RebusDrop takes over the remaining functionalities for seamless up- and downloading of the data as a background service in the windows tray.

This is much smoother than the former concept and is the next step to giving the user the sense of having a render farm right there on their desktop, rather than feeling they are sending the data and getting it back from a render farm service somewhere far away. Furthermore, this update comes with a bunch of other new improvements that you can read about on our homepage.
 

CGA: How many people work at RebusFarm and what are their roles?

RH: We have 10 employees at RebusFarm, all of them with strong 3D animation backgrounds. Five of them work as software engineers to improve the existing software and interface, keep things running, integrating more 3D software packages and improving the internal processes. Another five employees are in Customer Support, and they also work to improve the homepage, produce tutorials, solve issues for customers, etc.

CGA: Tell us about the infrastructure at RebusFarm and what was involved to set it up.

RH: Needless to say, the hardware infrastructure of RebusFarm is confidential. We customize our hardware extensively to reduce costs in a lot of ways, and also to fix problems in energy supply and thermal problems. If you think of 1500 Dual XEON machines in a limited space and the scale of electrical energy being converted into heat, then you can imagine the serious thermal issues you will soon have to face. If you said we produce nothing but hot air it wouldn’t be a lie, as the resulting images have no physical energy :) But the biggest problem we have is, How do we get the heat out of the Data Center?

CGA: RebusFarm appears to be quite committed to green energy and you say you are able to cool your server rooms without air conditioning. Could you tell us more about this?

RH: Yep. We are based in Germany and people are very "green-ish" here, so we are too. Using green power is a little bit more expensive than conventional electrical energy and electricity is not cheap in Germany anyway. But we think it is also a good sales pitch, so we didn’t raise our prices to cover the additional costs. In the end it was an ethical decision supported by sound business thinking.

Instead of using a conventional air conditioning system, the cooling at RebusFarm is done by evaporative cooling (adiabatic cooling). An evaporative cooler is basically a system that introduces water into the airflow. As the water evaporates, the air is chilled and pushed out into the room. The temperature can be controlled by adjusting the airflow of the cooler. Air conditioning consumes more energy to transport the heat out of the Data Center than the servers themselves consume. In other words using air conditioning raises energy consumption, doubling it or even more.

CGA: What are the biggest challenges in running a render farm?

RH: Right from the start we ran from one challenge to the next one and still do. As I said, you have to get so much energy into and so much heat out of the Data Center. You have to build hardware that can handle centrally a very large number of server nodes and distribute data to your customers efficiently. Another task is to develop a perfect interface between costumers and the render farm. This took years to develop and is still in progress. Another important and complicated point is the render farm management software in the Data Center. With each increase in the data we handle, it has to be adapted and expanded. Some people may think it is simply,Llet the machines run, but it is not.

CGA: What is the most rewarding part of running a render farm business?

RH: Well, if your addicted to something and you see your ideas come alive even after a long time of working on it then this is very satisfying. Also the apparently extensive approval in the 3D community is much appreciated by us. Sometimes clients show us their final projects and I am happy that we could help them, especially when the results are outstanding.

CGA: Running a successful render farm I presume is more than just a server room full of machines. What makes RebusFarm the company it is? Given there is so much competition, what makes one company better than another?

RH: That is really a question others should answer. We are sometimes obsessed with our project, but I think our competitors are too. In the end you will only enjoy success if you believe in your project and work hard, but that is nothing new. Also we have a free-minded and creative atmosphere among our people, encouraging new ideas and establishing a strong identification with our team (Or maybe I should say that is my hope!).

CGA: Tell us about some of your most recent changes to your service and where you plan to take the company over the next 5 years.

RH: We have seen a lot of new Render farm services hitting the market and also a lot of them leaving. But it is getting harder and harder for a new start-up to get into the market because of the head-start of the existing services. This involves the software and interface a service can offer to its clients, but also the price of the service. The larger a service is, the better are the possibilities to offer a good price to the client. That is self-evident. And RebusFarm is the the largest service available today.

CGA: The preliminary results of the CGarchitect industry survey show RebusFarm as the market leader by a considerable margin (around 9 times more market share than your next largest competitor). What are your thoughts on that and what you’ve done to get to this point?

RH: We monitor all activities in the CG Industry, especially in the field of render farm services, and I think we have been through a phase of consolidation in the last few years and this will continue in the future. It all depends on the ease-of-use of the tools, the service and the price you can offer. It is a competition of ideas.

CGA: What are your thoughts about GPU rendering and how this impacts future render farm services and the rendering industry for that matter?

RH: GPU Rendering is more common for Viewport rendering than it is for final frame rendering. It is not our main focus, but I cannot say what will come along in the future. If our customers ask for it then we will offer it.

CGA: What are your thoughts on cloud based rendering services and 3D applications as well, and real-time rendering and the future of the industry?
 

RH: I see cloud-based rendering growing as the demand for render capacity grows faster than the render power available for users in their work place. Recent progress in High Definition Television sets (HDTV, UHDTV) has been dramatic and we can only expect it to go on in this way, And with the spread of stereoscopic TVs, the rendering demand for those devices doubles the render load of ordinary 2D TV input.

There are ambitions to transform 3D Software from a local software to an SaaS (Software as a Service). I don’t think this will find its way into the mainstream. Apart from all the technical problems, 3D users always want to save the data on their local computer to protect their work and secure their privacy.