Quadrics QsNet provides both scalability to large numbers of nodes as well as balanced performance for large SMP nodes that is critical for such multi-TFLOP solutions. Quadrics QsNet will be used to unify all the processors within the Tera10 system to provide the necessary performance for computation and I/O. Tera10 will integrate 544 Bull NovaScale 6160 computing nodes, each including eight Itanium 2 processors.

The global configuration will feature 8704 processors with 27 terabytes of core memory. Each of these computing nodes will contain multiple Quadrics QsNetII (Elan4) network adapters. Beyond computational power, Tera10 will also house a huge storage capacity. The Tera10 configuration includes 54 additional Bull NovaScale I/O servers managing one petabytes (one million of billions bytes) of disc space with a sustained throughput of 100 GB/s. Each of these I/O servers also contains multiple Quadrics networks. Tera10 will operate the Bull HPC software platform that includes the Linux operating system and Lustre, the global and parallel file system. Quadrics will provide the Quadrics RMS software to support the scheduling and resource management for Tera10.

"Quadrics is very pleased that the QsNetII interconnect and software have been selected for the latest CEA supercomputer. We have developed an excellent relationship with Bull, resulting in a product that is unique both in terms of power and functionality", stated Quadrics' Chief Executive, Cristoforo Romanelli.

"Bull has worked closely with Quadrics to build clusters designed for high performance computing centres. These clusters encompass the power of Bull NovaScale servers with a unique scaleable implementation of Quadrics' networking technology. This combination will be able to provide other leading research centres access to massive computational and i/o resources meeting the demand for scaleable Supercomputing solutions", stated Gerard Roucairol, Bull's Chief Scientist.

Jean Gonnord, Numerical Simulation & IT programme director at CEA/DAM stated: "Our simulation programme requires outstanding performance from the network interconnecting the nodes of such a large machine, especially in terms of latency and barrier synchronization. Quadrics, that we already chose for our previous computer TERA-1, is still today the only one able to meet such a challenge."