Just before Christmas we launched a new body-measurement feature for our shapify.me project. A person goes into the 3D booth, gets scanned, and the body-measurements are calculated with the 3D scan data. We couldn't avoid using a GPU with these calculations.
How to install this branch of processing into our elegant scaling pipeline? When you're dealing with 3D scans you should remember two things: it's expensive and time-consuming. By the way, its workload fluctuates throughout the days. The easiest solution for us was to use Docker.
This speech is not about mining, but the most helpful articles for us were ones such as "Building your own mining farm". That's because we use quite a similar infrastructure.
GPU, Docker, scaling. What we achieved, where we failed, and what it became of it in the end is all included in my speech.