This release provides security patches and several enhancements:
Support for manually configuring pod attributes
Administrators can manually configure the attributes of pods in a hardware tier, including their resources, labels, and annotations. See Manually configure pod attributes in the Admin Guide for details.
Users can perform distributed CPU or GPU MPI-based training on Domino. Domino orchestrates Kubernetes level resource orchestration. This makes it easy to provision and use compute capacity for both batch and interactive workloads without DevOps overhead. Users also benefit from simplified code distribution to all cluster workers. This has been a usability issue in tradition MPI setups. Lastly, MPI clusters can use NVIDIA NCG distributed training images to make it easier to get started. See Open MPI for details.