home projects facilities staff publications datasets

Facilities


High Performance Computing

Research at the AMNH relies increasingly on high performance computing, with science applications extending from phylogenetic analyses of genomic and anatomical data in comparative biology to simulations of star cluster evolution in astrophysics (See Projects and Publications). Currently, the primary resources available for high performance computing consist of three systems.

Demeter: 128 nodes (including master) with two 2.8GHz XEON CPUs, 4GB RAM, and Myrinet and 100MB ethernet interconnect.

Eve: 8-way server with 2.2GHz Opteron 846 CPUs and 128GB RAM. (Shown here with Andrés Varón.)

Enyo: 33 nodes (including master) with two dual-core 3.0GHz Intel XEON 5160 CPUs, 16GB RAM, and Infiniband interconnect.

 GRAvity PipelinE (GRAPE)


(left to right) Mordecai MacLow (AMNH),Piet Hut (IAS), Mike Shara (AMNH), Simon Portegies-Zwart (OpenUniversity), Jun Makino (U. Tokyo)

Jarrod Hurley pictured with the GRAPE boards

The gravity pipeline, or GRAPE, boards used at the American Museum of Natural History were designed by astrophysicist Junichiro Makino of the University of Tokyo. These boards were designed to accelerate the solution of astrophysical problems involving gravitational interactions of large numbers of stars by moving the most computationally expensive parts of these problems from software to hardware. By encoding these algorithms into the hardware of the GRAPE boards, simulations involving thousands of star could be completed hundreds of times faster than if they were done on normal workstations. Because of the increase in speed problems which were previously intractable, such as those involving hundreds of thousands or millions of stars, are now possible. The Museum currently has five GRAPE boards which are used in a variety of astrophysical research programs (see Projects page).

 

SUN Cluster

The Museum's SUN cluster (generously donated by SUN Microsystems) consists of three 12-processor SunFire V1280 SMP machines and one four-processor V880z visualization server, all connected by a fast Myrinet network. The large number of processors per machine, the fast network connecting the machines, and the large amount of memory available (96GB in each V1280), makes certain problems more tractable, in particular those simulations which are required to follow very small details within very large datasets. With the visualization capabilities of the V880z, we are able to explore new ways to view and analyze these simulations.

 

SGI Onyx

Two SGI Onyx machines with 12 and 28 MIPS R12000 processors, 6 and 14 Gb of memory, and three and seven graphics pipes are the core of the Museum's efforts in scientific visualization. The Onyx, one of the largest supercomputers ever dedicated to creating visual simulations and equivalent in power to those used by NASA or the largest military research laboratories, serves as a major research tool for staff astrophysicists and is central to the Museum's Space Show in the new Hayden Planetarium.



AMNH | AMNH research | Astrophysics | Invertebrate Zoology | Earth & Planetary Sciences
Last Revised 2/2/2007 © American Museum of Natural History, 2003-2007