[an error occurred while processing this directive]

NERSC 3 Greenbook

Next: Parallel Vector Supercomputers Up: The Energy Research Computational Previous: The Energy Research Computational

Hardware

The hardware resources housed in the HPCAC serve the needs of both capacity and capability computing. Capacity computing engines provide high computational speed and rapid throughput for all but the largest jobs in the ER spectrum. Capability computing engines have special features (e.g. very large memory, very large local disk, high speed archive and data transfer engines, etc.) for jobs that require some combination of these features. In the following sections we discuss briefly the different classes of supercomputers in use for computational science and the supporting hardware infrastructure to make effective use of these supercomputers. It should be noted here that the definition of a ``supercomputer'' is a very fluid and time dependent definition.

The HPCAC user base has it's primary experience with parallel vector supercomputers, but usage of RISC-based symmetric multiprocessors (SMP) systems and of massively parallel processing (MPP) systems is increasing. NERSC has offered its own MPP (A CRI/SGI T3E) since 1996, and now offers a mix of MPP and PVP hardware; a RISC-based SMP is scheduled to come on-line at NERSC in the Fall of 1998. The current and expected utilization and the experience gained will determine the hardware and software requirements for the next major expansion of the facility.

NERSC 3 Greenbook

Next: Parallel Vector Supercomputers Up: The Energy Research Computational Previous: The Energy Research Computational