Hardware

(Updated April 2020)

The HPC Infrastructure 

Cineca is currently one of the Large Scale Facilities in Europe and it is a PRACE Tier-0 hosting site.

• MARCONI: It is the Tier-0 system that replaced FERMI in July 2016. It is based on the LENOVO NeXtScale platform and the next generation of the Intel Xeon Phi product family. It has been gradually upgraded from June 2016. The actual configuration consists of Marconi-A3 with SkyLake (in production since August 2017, upgraded in January 2018 and completed in November 2018). In a previous configuration phase, Marconi had two additional partitions: Marconi-A1 with Intel Broadwell, into production since July 2016, closed in September 2018, and Marconi-A2 with Intel KnightsLanding into production since January 2017, closed in January 2020. Marconi is classified in Top500 list among the most powerful supercomputer:  rank 12 in November 2016, and rank 19 in the November 2018 list.
• MARCONI100:  It is the new accelerated not conventional Marconi partition available from April 2020. This is an IBM system equipped with NVIDIA Volta V100GPUs, opening the way to the accelerated pre-exascale Leonardo supercomputer.
• DGX: It is the NVIDIA A100 accelerated system available from January 2021, particulary suitable for Deep Learning frameworks. This is a AMD 3-node system equipped with 8 NVIDIA A100 Tensor Core GPUs per node. 
• GALILEO: It has been renewed in March 2018 with Intel Xeon E5-2697 v4 (Broadwell) nodes, available for the Italian research community.
• CLOUD.HPC: CINECA cloud service has been renewed in April 2020 with Intel E5-2697 v4 (Broadwell) nodes, completing the HPC ecosystem. 

The Data Storage Facility

• Scratch: each system has its own local scratch area (pointed by $CINECA_SCRATCH env variable)
• Work: working storage is mounted to the three systems (pointed by $WORK env variable)
• DRes: a shared storage area is mounted on all machine's login-nodes and all Pico's computes node (pointed by $DRES env variable)
• Tape: a tape library (12 PB, expandable to 16PB) is connected to the DRES storage area as a multi-level archive (via LTFS)

Old HPC Infrastructure

• D.A.V.I.D.E.: (Development of an Added Value Infrastructure Designed in Europe)  It was the energy-aware, High-Performance Cluster, based on OpenPOWER8 servers and NVIDIA Tesla P100 data center GPUs.  It entered the Top500 and Green500 lists in June 2017. D.A.V.I.D.E. was in full production from January 2018 to January 2020.
• FERMI: it has been a Tire-0 system from June 2012 to July 2016 and gets its name from the famous Italian physicist. It is an IBM BG/Q supercomputer, classified among the most powerful supercomputers in the Top500 List: rank 7th in June 2012. 
• In June 2012 it was ranked 11th in the Green500's energy-efficient supercomputers list.
  FERMI was taken out of production as July 18, 2016, substituted by the MARCONI system.
• GALILEO: It is the second system (Tier-1), it is an IBM NeXtScale cluster accelerated with Intel Phis. Galileo replaced the old PLX system. It is named after the Italian physicist and philosopher who played a major role in the scientific revolution during the Renaissance. Galileo was in full production from February 2015 to November, 20th, 2017.
• PICO: It was made of an Intel NeXtScale server, designed to optimize density and performance, driving a large data repository shared among all the HPC systems in CINECA. It was used for "BigData" classes of applications, complying with the peculiar hardware requirements (large memory per node, massive storage equipment and sharing, fast data access, and transfer, etc.), the software tools and the high-throughput technologies needed by data-oriented projects, such as an accelerated visualization environment, cloud computing, hadoop etc.