Before we started our project, we identified the goals
and accomplishments to be achieved by the project.

We broke our project of creating “Animux Caseless Animation Rackmount Renderfarm” down in to 3 major groups which were itemized so we could place a clear accomplishment for each group.

1. DESIGN
• Attainable: By using material that is easily available to build the renderfarm, we were able to accomplish this task. As will be evident in our article, you will notice that everything we used in building the rackmount is already available in pre-built form. You will only need simple household tools to built it. This decision caused the longest delay, because of the challenges posed by this restriction as the people on the project were not handymen. However, we are happy that we did it, as we now, you will not need them either. (If my 60+ [computer-unsavvy, relative in-experienced with tools] year old parents can put one together, I think it is safe to assume you guys can too.)
• Portable: Our plan was to demo this Rackmount at our Siggraph or other animation festival presentation. For this it has to we should be able to fit atleast 32 CPU core renderfarm in a single suitcase. I think this is quite achievable. We also knew most of the people who would try it out would have to move it from place to place within their studio. This we have definitely achieved it in the past 3 months we have moved the renderfarm around the basement quite a few times. With the wheels on the rack we could move it EFFORTLESSLY, I might add.
• Expandable: People familiar with the task of setting up/using renderfarms know, that it is only a matter of time you will have to expand your farm. But this should be a simple process, if you have a process to build the next set of render nodes. We started our project in the CAD design software to design out Racks by guys who build things far more complex than a rack. Came up with a template with all the measurements and just following the process to create more nodes for the rackmounts. Also, using the IKEA stand and shelves (whose measurements will never change) helps with now having a stable process to expand.
• Efficient: Here we are talking about efficiently using the volumetric space for our rackmount. The goal is to put in as many motherboards, CPU’s as safely possible into our IKEA wireframe shelf. As we decided to go with a higher end motherboard for our GPU based render pipeline, it was a challenge. We also didnot want to compromise the strengh of the wireframe by cutting a few wires out (I tell you we were tempted many a times!!!). But we went back to the CAD program and tried various configurations to finally come up with one that worked. You will also notice that we found efficiencies in building the supports for the motherboards by using fewer paint stirrers and lesser cuts in our final version than the previous 2 versions.
• Durable: All the above goals would be pointless if the set up was not durable. We took extra precautions to make sure that we figure out the weakest link and address it. Finding a place to store the powersupply was what we found was the weakest link. We have addressed it in this version which is pretty stable. For the past 3 months things are still functional with no problems, which makes us happy with its durability. We are being our usual selves working on the next version to make it better than what we have today.
• Cheaper than packaged Hardware: Meeting all the above specifications and requirements while being the cheapest way to get the same amount of computing power makes our solution the appealing. One requirement, we always sought after was to spend almost all the money on the computing power and as less money as possible on the non-computing components while meeting the above design requirements.
2. FUNCTIONAL
• Maintainable: One of the critical goals was that the renderfarm should be easy to maintain. Which meant, we should be able to repair on a single rack without having to bring down the entire rackmount. That meant, we needed to set all our connections out to the side or front. With the existing setup we were able to repair, upgrade or replace a single rack while the Animux Caseless Rackmount Animation Renderfarm was still rendering. We were also able to just work on a single motherboard while on the rack. (It was a little tight, as we had to leave the connections on the other motherboard, but it was do able. We however, would recommend taking the rack out completely.)
• Efficient: It has to energy efficient. One way to do that was to reduce the number of fans needed to cool the rackmount. In order to achieve this, we right from the beginning decide to make the rackmount caseless. The power supplies are set for each rack to pull the hot air from inside (in the middle) and push it out in the front. We plan to use only one chasis fan to push the hot air towards the power supply fan for proper cooling. We rendered the jobs with the Animux OS with the absence of hard-disk on the slaves. This futher improves our energy efficiency.
• Performance comparable to Off-the-Shelf Rackmount: We were able to render a scene (400 images) out of Bassam Kurdali’s “Mancandy FAQ” in 20 minutes on the Animux Renderfarm without tweaking the OS, just by booting into the Animux OS renderfarm options. When we first rendered this on a pre-built machines used as nodes it was 2 hours or more. (However, these are using quad cores and the pre-built was a single core.Also, the Animux Caseless Rackmount Animation Renderfarm slaves had no harddisk and booted out of the Animux LiveUSB to complete the render.) We plan to add more tests with numbers in our “Metrics” portion of the article page.
• Stable: The fundamental functional goal is for Animux Caseless Rackmount Animation Renderfarm to be stable. It has to boot everytime we turn it on without any fiddling. Once it is up, it should stay up until the render task is complete. Our preliminary tests have proved it is stable setting. We plan to further stress test as the days go by and are quite confident it will live up to the challenge.
3. FUTURE PROOF
• Easily Upgradable (Scale Up/Scale Out): Because Animux caseless rackmount animation renderfarm is easy to maintain it is easy to upgrade (scale up) it. Because it was designed to be expandable (scale out) and cheap, to create another rack would cost you $14 for the stand and $20 for the racks in non-computing power expenses. You build another of similar rackmount tailor made for your need.
• Current with (& Ahead off) Existing Technology : Rackmount had to be future proof while being current. I think we did the best we could in our goal to get a 48 CPU core renderfarm for under $3000. We had to pick motherboards that can atleast handle an upgrade. (We could have the best MB, but because of budget constraint, we decided on our choice which we think is good enough.) We also wish to use GPU based rendering, which though in its infancy is something we right from the beginning of the project wished to use.

Back to the Main Article

Tags: 3D, 3d animation, Animation, Animation Renderfarm, blender, how to build renderfarm, rackmount, Render, Render Farm

This entry was posted on Friday, May 8th, 2009 at 11:45 am and is filed under featured, Render Farm. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.