The client had a requirement to implement a cluster of CUDA GPU servers in the DigitalOcean cloud to provide them with rendering the blender file for output.
The client had a timeline for us to get this done as they had the footage output of the blender to provided in a mission-critical environment. The process was also bonded to a budget and has to cope with the same. Since load was high, the process has to split across different instances and joined together at the output.
Skynats research get in place as soon as the project was in place. We swiftly checked through the cloud platform and tested their GPU instances running Tesla Nvidia. Apparently, this GPU was optimized for the AI and provided less output for image rendering, hence we found a better solution in CPU based render of the blender. The cloud platform was chosen and the data was split according to the frames in the blender in the project (with the help of client), with which we scheduled jobs for each node in the cluster and point out the result to common storage. Since the process was running simultaneously in different nodes, the whole process took one-tenth of the time needed in the actual process. The process was implanted at the time for the delver of the project. We also created a template so that they can use this again for the next projects to come.