You are here


Storage and processing of large volumes of data

PICO is the new system of Cineca, for Storage and processing of large volumes of data. It is named after Giovanni Pico della Mirandola, humanist and philosopher renowned for his prodigious memory, who lived in Modena, near Bologna, the 1400s.

The new system is intended to enable new "BigData" classes of applications, related to the management and processing of large quantities of data, coming both from simulations and experiments. Particular reference is put on all emerging scenarios as interactive model (Urgent Computing model), as well as new paradigms of resource utilization through "Cloud Computing" technology. The coexistence of virtualization technologies (cloud computing), orchestration technologies (access on demand) and high performance computing (Tier-0 and Tier-1 systems), completes the data-centric architecture of the HPC infrastracture of Cineca.

PICO is 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.

The storage area is composed by hight throughput disks (based on GSS technology) for a total amount of about 4 PB, connected with a large capacity tape library for a total actual amount of 12 PByte (expandible to 16 PByte). The storage area, accessible from all HPC systems in Cineca, is a "multi-level" memory.

Model:  IBM NeXtScale

  Total Nodes  CPU Cores per Nodes Memory (RAM) Notes
Compute/login  node 66 Intel Xeon E5 2670 v2 @2.5Ghz 20 128 GB  
Visualization node 2 Intel Xeon E5 2670 v2 @ 2.5Ghz 20 128 GB 2 GPU Nvidia K40
Big Mem node 2 Intel Xeon E5 2650 v2 @ 2.6 Ghz 16 512 GB 1 GPU Nvidia K20
BigInsight node 4 Intel Xeon E5 2650 v2 @ 2.6 Ghz 16 64 GB 32TB of local disk

Data storage, Visualization and Data Analysis

Simulations running on modern HPC machines can produce large amounts of scientific data, that sometimes need to be post-processed with data analytics and visualization tools. Starting from mid 2014 CINECA provides a facility to better face these kind of applications. This system, called PICO, consists of a smaller cluster of standard processors, with a higher amount of central memory and accelerators for enhance the rendering process. PICO also has faster direct access to the shared data repository to allow fast reading of data from simulations.

PICO is a BigData system specifically tailored for data-analysis and visualization. A large and fast data repository (nearly 4 PB, called $DRES) is connected to PICO and accessible in a shared way with all the HPC systems. It is intended for sharing data within the project team and across computational platforms. $DRES is a multi-level storage via LTFS.  A new Tape library is connected to PICO, equipped with 12 PB (expandible to 16PB) of space for uncompressed data. It is available only though the $DRES filesystem with the LTFS technology. A specific entry is defined for each username ($TAPE)  in a shared way across all HPC systems, for private archiving of data on magnetic support for long time archive.