We are proudly releasing a nice update to our microservices cluster!
We recently added shared GPU support on top of Kubernetes. This was needed to deploy some larger deep learning models. The task was not easy as, by default, Kubernetes schedules only one container per GPU, which results in a considerable waste of computing resources. To avoid that, we used a GPU sharing scheduler to share a GPU between multiple containers.
We also have set up some nice demonstrators of machine learning microservices here: https://icosys.ch/icoservices