HPC | Ricerc@Sapienza

High-Performance Computing with TeraStat

Abstract—In the last years we are witnessing a progressive
shift toward the adoption of the cloud-computing model for
a large variety of application domains. High-performance
computing (HPC) is not an exception to this trend, with
a steadily increasing amount of performance-heavy scientific
experiments that are conducted on the cloud. However, this
approach may pose some serious security, operational and
economical issues.
In this paper we report about our experience with this

On power capping and performance optimization of multithreaded applications

ulti-threaded applications facilitate the exploitation of the computing power of multicore architectures. On the other hand, these applications can become extremely energy-intensive, in contrast with the need for limiting the energy usage of computing systems.

SapienzaAI

Lingua Italiano

Dual NVIDIA DGX-1 AI System providing a 2 petaFLOPS computing power optimised for Deep Learning. The two machines are connected with IB EDR high performance network.
System specs:
2 x NVIDIA DGX-1 systems, each one cnsisting in 8 NVIDIA® Tesla V100 GPUs with 256 GB total GPU-RAM, 40960 CUDA cores, 5120 Tenosr cores, 512 GB system memory, 8 TB storage, 4 IB EDR

Monolith

Lingua Italiano

Monolith è il cluster di supercalcolo del Dipartimento di Ingegneria Meccanica e Aerospaziale per applicazioni di calcolo parallelo (HPC).

Monolith è un sistema per calcolo ad alte prestazioni ottimizzato per lo sviluppo di applicazioni di calcolo parallelo (HPC) con l’opzione di acesso a GPU di alte prestazioni.

Monolith è stato di recente espanso grazie al contributo di fondi per Dipartimenti di Eccellenza.

Il sistema nella versione espansa è costituito da:

TeraStat 2

Lingua Italiano

TeraStat2 is an HPC infrastructure developed by the Dipartimento of Scienze Statistiche and hosted by the InfoSapienza IT center of University of Rome - La Sapienza. It provides a general-purpose, massively parallel supercomputing infrastructure for solving large mathematical models on Big Data.