The Department of Computer, Control and Management Engineering (DIAG) at the Sapienza University of Rome was established in 1983 as a university research and training center and carries out advanced multidisciplinary research in control engineering, management engineering, computer engineering, bioengineering and operations research. In 2001 it was named after Antonio Ruberti, the eminent scholar who founded it.
As part of the “Departments of Excellence” initiative, the Ministry of Education, University and Research (MIUR) awarded the 180 best Italian university departments with an extraordinary contribution to finance five-year development projects. Among these, DIAG was selected thanks to the project on securing cyberspace: the funding obtained was also allocated for new equipment for its research groups, in particular Cybersecurity, Artificial Intelligence, Smart Environments , Natural Language Processing, Data Science and Analytics.
DIAG needed a flexible tool with computational power for advanced Deep Learning research applications: a platform capable of allowing the prototyping, training and testing of solutions based on deep neural networks, with the possibility of multi-user use. DIAG’s goal was to create a sort of private cloud for the Department for Machine Learning applications with a containerized structure. The solution had to include suitable software with the possibility of using Python notebooks by multiple users simultaneously, ensuring tight isolation of the software environments used by the various users via Docker containers.
beanTech proposed NVIDIA DGX A100, the universal system for all AI workloads.
NVIDIA DGX A100 delivers unprecedented compute density, performance, and flexibility. The DIAG configuration packs 5 petaFLOPS of AI performance into 6 rack units, thus replacing traditional computing infrastructures with a single unified system.
With NVIDIA DGX it is possible to allocate compute resources in a granular way, using the Multi Instance GPU (MIG) feature introduced with the Nvidia A100 GPU. This allows system administrators to allocate the correct number of resources to a specific workload.
The DGX system features 8 Nvidia A100 GPUs with up to 640GB of overall GPU memory that boost performance on larger training runs.
THE SUCCESS STORY AT A CLICK
Discover all the features and benefits of the solution created together with NVIDIA and DIAG from the Sapienza University of Rome. *This document is in italian version*
Having the requirements of the academic and research world understood, and having found a supplier able to meet those requirements rapidly, in an extremely professional and competent manner, has been an enrichment for us too. Research methods often have different needs to those of a company, and often the supplier is unable to perceive and understand them immediately. beanTech on the contrary managed to support our Department AI infrastructure innovation project in an exceptional way. The hope is that the experience can be repeated, since AI and specific infrastructures evolve over time we too must keep up.