This IBM® Redpaper publication describes the architecture as well as the installation procedure and the results for running a typical training application working on an automotive dataset, in an orchestrated and secured environment, providing horizontal scalability of GPU resources across physical node boundaries for deep neural networks (DNN) workloads. This Redpaper is mostly relevant for system engineers, system administrators, or system architects responsible for data center infrastructure management and typical day-to-day operations such as system monitoring, operational control, asset management, and security audits.