As a startup, we entered the market with our nationally innovative technology for rendering process automation using dedicated software we had developed. This software consisted of two basic modules, an uploader running on the client side, and a queuer running on our render farm side.
Both of these very powerful programmes, or rather a chain of interconnected programmes and scripts, were created to automate the rendering process on the render farm. What is an automated render farm? Simply, it is a hardware and software solution that allows an external authorised user to render their scene using dedicated computing power. Under ideal conditions, such a client can carry out the entire task without even a single contact or intervention with the staff operating such a farm.
Such an ideal situation is very rare, but achievable. It requires the development of a number of IT solutions that perform the following tasks:
- They will transfer the 3D scene together with all the necessary files and assets from the client computer to the computing cluster. It is worth mentioning here that this seemingly simple activity means finding all the files linked by the client, packing them and then unpacking on our computing cluster and relinking all the files to their current storage location.
- Specifying the material to be rendered including frame range, resolution, output file types.
- Distribution of the frame range as computational tasks to the computating nodes (servers) assigned to the task.
All this must be carried out with full compatibility between the software, the rendering engine and all plug-ins used by the client.
With our solution, the client installs our plugin for their software (for example, 3D Studio Max) and then runs the rendering on our farm using the plugin.
Our system then allocates computational resources according to the client’s priority and hardware requirements, usually following the principle of one animation frame being rendered by one node.
This all seems trivial, but it took more than two years of developers’ team hard work to create this solution.
And when we were ready we launched as a new service startup.
What did automation give us? Well, thanks to automatic rendering, we were able to increase the capacity of customers 10 times. It should be noted here that all the above-mentioned activities previously had to be performed by a human being. Moreover, these activities are mechanical, repetitive and subject to a high error risk. But most importantly, they were time-consuming and this was reflected in the number of customers that could be served.
However, after the introduction of the automated system, this bottleneck was eliminated and another one appeared, namely the available computing power. The idea behind our service was to make it as scalable as possible so that an increase in the number of customers would not overload our staff and slow down the service. However, there was still the aspect of the amount of computing power we could offer. This meant we had to make a decision, which was essentially a choice between taking the easy way out, making us dependent on external, IT giants, or trying to realise our own vision for the company while maintaining our independence. The computing power market offers the possibility of renting servers, including computing servers, in almost unlimited numbers. We could, like some of our competitors, instead of making our own hardware investments, rent them from giants such as Google or Amazon. This would have meant reducing our service to a kind of ‘overlay’ on the offerings of these entities. We could rent the servers we used to provide our service from these entities. In such an arrangement, we would not have had to invest more in hardware and could have focused solely on our service.
Another solution was to constantly expand our hardware base so that we could serve more and more customers.
And it was that moment when all the support from the European Union kicked in, a project aimed at almost doubling our computing power. It was a period of rapid growth and very hard work for our company, but thanks to this project called: “Fully automatic rendering for Polish and foreign large business customers, including film studios”, we were able to start serving large entities. The money was substantial, as the value of the project exceeded eight million zloty. It was therefore ten times more than the funds we had received from venture capital to start up just two years earlier.
This time we were smarter… a little bit. We opted for homogeneous computing servers, on the premise that we needed a small cluster of very fast servers to serve business clients and a large number of slower servers for less demanding clients. Unfortunately, we are not allowed to boast about which companies and films we have served, but I think that everyone in Poland has seen at least two or three animations that have been created on our servers. And not only in Poland, as the domestic market proved to be too small for us. This is why we placed particular emphasis on the internationalisation of our business, working with numerous resellers abroad, with the Korean market being the most important. We are, of course, talking about South Korea, where, as it turned out, no competing render farm existed at the time, and the concept of outsourcing power itself was surprising. Therefore, if a k-pop fan is reading this, they can easily guess who rendered the special effects for their favourite music videos.
Interestingly, and perfectly defining the speed of technological progress, this month we are recycling the 360 computing servers we bought as part of this project. Technological progress in IT is still very fast and Moore’s Law, which states that the computers computing power will double every 18 months, still holds true, at least for supercomputers.
The project has been a great success, and has allowed us to put ourselves on an equal footing with the largest providers of similar services around the world.
At the same time, it was this project that made us realise the importance of constantly undertaking R&D in such a highly competitive environment as rendering services on an automated render farm. Thanks to this project, we realised that the computing power dedicated to rendering alone, even if it is the largest resource of computing power in Poland, is not enough for the constant and reliable development of the company.
