CLOUD

NASA and Altair continue their collaboration as NAS expands its HPC capacity to 700 teraFLOPS in 2008: Altair Engineering, Inc., a leading global provider of technology and services that empower innovation and decision-making, today announced that the NASA Advanced Supercomputing (NAS) Division at NASA Ames Research Center, Moffett Field, Calif., has entrusted workload management of its huge new Pleiades cluster to PBS Professional software from its GridWorks business unit. Pleiades, an SGI ICE cluster with 51,200 cores across 12,800 Intel Xeon processors and 51TB of memory, gives NAS 609 teraFLOPS of peak processing power. A recent LINPACK benchmarking projects its placement at No. 3 on the Top500 list of the world's fastest supercomputing systems. The unique scalability of PBS Professional enables NAS to run a single job across all Pleiades nodes from one PBS server. "PBS Professional does just as good a job managing this big cluster as it does managing our Columbia supercluster. All of Columbia's systems use Itanium processors with Linux (single-system image (SSI) systems, 18 with 512-cores, 2 with 1,024 cores, and one with 2,048 cores)," said Alan Powers, HPC Technical Director, CSC. Pleiades increases total teraFLOPS for NAS HPC capacity by more than 7x over pre-installation capacity. It will be used by NASA scientists and engineers to perform simulations for agency missions, such as spacecraft design, and to expand knowledge of the universe by modeling celestial bodies and phenomena such as merging black holes. Pleiades, with 512 Xeon cores in each of 100 racks, is the largest InfiniBand-equipped system ever built. Its double data rate (DDR) dual fabrics require over 20 miles of InfiniBand cable. Compute nodes can communicate over either fabric for greatest efficiency. PBS Professional manages the scheduling of activities on all NAS computing resources, including Columbia, an IBM 590 system named Schirra, NAS archive post-processing servers, and hyperwall-2, a tiled display with 128 20-inch LCD panels. Each of the hyperwall's 128 nodes includes a dual socket quad-core Opteron processor. "NASA has used PBS Professional since 2000 and used PBS long before then," said GridWorks CTO Bill Nitzberg. "So we're very pleased to be working with NAS on management of Pleiades and to continue working with them as we develop management technology for next-generation HPC systems." Over its long history, Altair has been the only workload management company to make the kind of investment in scalability that now enables NAS to manage a cluster of Pleiades' size from a single server. As NAS HPC resources grow, Altair will work with them to ensure that GridWorks workload management technology still scales to meet their needs and those of supercomputing sites around the world.