Laboratory Resources
High Performance Computing resources will be mainly provided at CILEA site.
Some of the most advanced vector and parallel computers are already available at CILEA.
1) 100 PA-Risc CPUs HP Hyperplex cluster (64bits technology). High speed interconnection (Full-duplex 320MBytes/sec bandwith). 100 Gbytes RAM and 2TBytes disk storage. Peak performance: 240GFLOPS/s.
2) 4 CPUs NEC vector computer, 16 GByte RAM and 250 GBytes disk storage. Peak performance: 16GFLOPS/s.
3)128 nodes Eurotech SpA cluster computer ( installed January 2004) using 256 INTEL XEON at 3,06 GHz. 512 GByte RAM and 11 TByte disk storage. Connection between nodes: Myrinet 2000 2 Gb/s full duplex (computational); Gigabit Ethernet (management and distributed file-serving). Peak performance: 1,57 TFLOPS/s, Sustained performance 1,08 TFLOPS/s The cluster was upgraded in April 2005 with 18 4-ways AMD Opteron 2,2 GHz with Infiniband 4X interconnection, for a total peak performance od 1,88 TFLOPS/s
4)Three HP computers Intel Itanium processor based. (64 bits) These are addressed to analysis and development performance of grid-based technologies.
The computing service offered by CILEA is supported with additional resources that contribute to increasing the reliability level and quality of the service:
The computer room (>500 square meters) with controlled entrances, fully conditioned and independent power supply. This is the room that will host the Supercomputer purposely designed on behalf of the Interdisciplinary Laboratory of Bioinformatics Technologies.
In addition to the preexisting resources a new dedicated supercomputer will be installed.
The computing platform best suited to the needs of the Laboratory is a cluster computer, based on 4-ways nodes, 64bits CPUs (at least 256 CPUs) with high speed and performance internal network.
The decision to choose 4-ways nodes is a compromise between the need of having dense and low-cost modules and the advantage and availability of high speed interconnections and a large amount of central memory (RAM) used by the parallel threads of the computational jobs.
Besides the low latency will optimize the execution of parallel jobs. The peak sustained performance will be at least 1.2 TFLOPS.
The backup and storage system will be as follows:
a) ADIC i2000 library, 4 LTO 2 drives and 600 slot that can host Ultrium 2 200/400 GB cartridges with a theorethical capacity of 240 TB; b) EMC2 CLARiiON FC4700 Storage Area Network System, twin controllers, 6TBytes on line disk space, that can be expanded over 25 TB.
The users of the new Laboratory will access all the described resources using a scheduled system that will be installed on the computing platform of the Laboratory. Besides WEB based interfaces will be available to access the Laboratory and other resources using GRID technologies on wide area networks.
GRID
The computing resources wiil be also accessible through GRID technology applications. The GRID platform will be based on the ones implemented in GRID-IT project (M. Vanneschi) and in EGEE project (M.Mazzuccato INFN).
The nodes of the GRID that will be implemented for this project will be at the premises of the Project Units (CNR di Genova, Milano, Napoli; IST e CBA Genova, Università di Milano, Genova, Camerino, Cagliari, Napoli; and other national centres (CINECA, CASPUR, CEINGE) and others ones at international premises.
