High-Performance Computing for Artificial Intelligence at the University of Turin
The University of Turin has created an open-access laboratory on High-Performance Computing (HPC) for Artificial Intelligence (AI). A centre capable of collaborating with entrepreneurs to boost their ability to innovate on data-driven technologies and applications.
Designed to foster research collaborations
We are open to collaborations in research projects (EU, National, Regional). HPC4Ai currently support 10+ EU research project in different research areas, and it is the integration platform for the whole ADMIRE EuroHPC project. The Computer Science Department, with its 90+ professors, covers the major research areas: HPC, AI (Deep/Federated Learning, Reasoning, Perception, …), BigData, UI/UX, Blockchain, Algorithms, Edge/IoT.
Computing support for research projects and pre-commercial solutions
HPC4AI implements a cloud-HPC system targeting AI workloads. Designed for research, allowing end-to-end configuration options: from HW to OS to cloud to service. The portfolio of services includes VMs+Storage, virtual clusters with access to OpenStack API (IaaS & PaaS). Ready to manage critical data (NGS, medical imagining, radiomics, finance, etc.).
Robust production system with fair pricing and support for experimental systems
HPC4AI cloud production system has been running for 3 years with zero incidents due to a robust configuration based on a self-stabilising automata that manage the system. A twin system called Playground is dedicated to developing new solutions, which are moved to production after extensive testing. HPC4AI playground is also the UNITO’s zoo for exotic computing systems: RISC-V systems and accelerators, novel Arm systems such as NVidia dev-kit and (when it will available) the EU-designed Rhea processor from the European Processor Initiative SGA.
Tier-III green datacentre (1.1 PUE) for convergent computing
HPC4AI is hosted in a self-designed 250KVA 16 racks Tier-III adiabatic green data centre, running over 90% of energy efficiency. See the HPC4AI real-time energy monitor (load, energy, PUE, etc.) and the environmental monitor (temperature, humidity, VOC, PM10/2.5, CO2, NH3, etc.). HPC4AI review the traditional usage of cloud and HPC systems to serve AI applications. The cloud provides a modern interface for HPC, and HPC serves as an accelerator for the cloud. A novel 2-phase cooling server (CPU+GPU) is currently under development.
GPU-enabled multi-tenant cloud with HPC4AI’s unique management tools
An OpenStack cloud system exploiting over 2400 physical cores, 60 TB RAM, 120 GPUs (NVidia T4/V100/A40), 25 Gb/s networking and 4 storage classes with different characteristics and costs. HPC4AI run UNITO’s designed management tools, such as the OpenDeepHealth PaaS to deploy private/secure/elastic instances of Kubernetes extended with single sign-on with identity and access management (keycloack), and VPN and firewall management (pfsense). OpenDeepHealth has been selected by the EU innovation program 2023.
Modular HPC system with UrgentSLURM
A modular HPC system with 68 nodes Intel (32 cores, 128GB RAM per node, OPA 100Gb/s), 4 nodes Arm (Ampere Altra 80 cores, 512GB RAM, 2xA100, GPUs, 2xBF2 DPUs, IB 100Gb/s), 4 nodes Intel (40 cores, 1 TB RAM, T4+V100 GPUs, IB 56Gb/s), 2 HPC storage systems (BeeGFS and LUSTRE all-flash). The system is managed with HPC4AI UrgentSLUM, a self-developed SLURM extension to manage the booking of nodes via a web calendar.
Designed for research: experimental systems and award-winning tools
HPC4AI provides developers with experimental systems, such SiFive RISCV servers, Esperanto RISC-V boards, NVidia dev-kit with Bluefield DPUs, as with novel tools for converged computing cloud-HPC that are the foreground technology of UNITO’s research, such as the StreamFlow WMS, and Jupyter-workflow co-lab. Streamflow has been selected by the EU innovation program 2023.