June 2, 2011

The Weekly Top Five

Tiffany Trader

The Weekly Top Five features the five biggest HPC stories of the week, condensed for your reading pleasure. This week, we cover NERSC’s acceptance of its first petascale supercomputer, the potential for magnets to revolutionize computing; NCSA’s private sector supercomputer; the official debut of Australia’s MASSIVE supercomputer; and PRACE’s biggest supercomputing allocation yet.

NERSC Accepts ‘Hopper’ Supercomputer

The National Energy Research Scientific Computing Center (NERSC) has officially accepted its first petascale supercomputer. The Cray XE6 system was named “Hopper” in honor of the renowned American computer scientist Grace Murray Hopper. The supercomputer will benefit more than 4,000 researchers and will support advancements in the fields of wind energy, extreme weather, and materials science.

NERSC Director Kathy Yelick, commented on this latest achievement:

“We are very excited to make this unique petascale capability available to our users, who are working on some of the most important problems facing the scientific community and the world. With its 12-core AMD processor chips, the system reflects an aggressive step forward in the industry-wide trend toward increasing the core counts, combined with the latest innovations in high-speed networking from Cray. The result is a powerful instrument for science. Our goal at NERSC is to maximize performance across a broad set of applications, and by our metric, the addition of Hopper represents an impressive five-fold increase in the application capability of NERSC.”

NERSC is the U.S. Department of Energy’s primary high-performance computing facility for scientific research. A pictorial journey of the delivery and installation process can be found here.

Chameleon Magnets Hailed as Potential Game Changers

Researchers at the University at Buffalo (UB) are studying the behavior of magnets and exploring their potential to revolutionize the field of computing. The researchers are asking questions about the nature of magnets and whether it’s possible to control their behavior to create more versatile transistors.

In the current issue of Science, University at Buffalo researcher Igor Zutic, a theoretical physicist, together with fellow UB physicist John Cerne, discuss the results of a Japanese study that demonstrates the potential to turn a material’s magnetism on and off at room temperature.

The release explains the basis for the research:

A material’s magnetism is determined by a property all electrons possess: something called “spin.” Electrons can have an “up” or “down” spin, and a material is magnetic when most of its electrons possess the same spin. Individual spins are akin to tiny bar magnets, which have north and south poles.

Zutic explains that the ability to switch a magnet “on” or “off” is revolutionary, bringing with it the promise of magnet- or spin-based computing technology — called “spintronics.” Spintronics-based devices will store and process data by exploiting electrons’ “up” and “down” spins. These spin states are similar to the ones and zeros found in standard digital transmission, but the technology makes it possible for more data to be stored using less energy.

Chameleon magnets could set the stage for a new era in processor design, and according to the researchers, may one day bring about the “seamless integration of memory and logic by providing smart hardware that can be dynamically reprogrammed for optimal performance of a specific task.”

NCSA Brings Supercomputing to Industry with iForge

The National Center for Supercomputing Applications (NCSA) is launching a supercomputer, called iForge, which will be dedicated to the center’s industrial partners. Rolls-Royce, Boeing, and Caterpillar are few of the companies that will be putting this computer cycles to work on a range of modeling and simulation problems.

A 22-teraflop high-performance computing cluster, iForge employs 121 Dell servers and a mix of Intel Xeon AMD Opteron processors designed to optimize workflows. 48 cores worth of high-level AMD parts are on hand to support memory-intensive pre- and post-processing jobs and highly-threaded applications. The system’s nodes are connected with 40 gigabit QDR InfiniBand from Mellanox. iForge doubles as a Linux-cluster or a Windows machine, since it runs both Red Hat Enterprise Linux and Windows HPC Server 2008 R2 operating systems.

In a prepared statement, Merle Giles, director of NCSA’s Private Sector Program, comments:

“iForge is a unique resource at NCSA, as it is designed specifically for commercial and open-source applications widely used by industry. This machine offers our Private Sector Partners several platforms to reach higher and higher levels of scaling and performance for physics-based modeling and simulation applications.”

More information about NCSA’s Private Sector Program is available at industry.ncsa.illinois.edu.

Australia’s MASSIVE Supercomputer Opens for General Use

Australia’s MASSIVE (Multi-modal Australian ScienceS Imaging and Visualisation Environment) supercomputer is now open for general use. The resource is part of a collaboration that includes the Victorian Partnership for Advanced Computing (VPAC), the Australian Synchrotron, CSIRO, Monash University, and the NCI. The State Government of Victoria also provided funding for the project.

The MASSIVE supercomputer is comprised of two tightly-coupled high performance computers — two 42 node IBM iDataPlex systems, each having 84 NVIDIA M2070 GPUs, 504 Intel Westmere compute cores, and 2 TB of memory. The combined resource offers 1,008 CPU-cores and 168 NVIDIA M2070 GPUs. Ten nodes have been upgraded to advanced M2070Q GPUs and 192 GB memory each, to address the specific requirements of interactive visualization workloads. Each system uses a high performance GPFS parallel file system, and both Linux and Windows HPC Cluster-based services are available.

The allocation process is open to the Australian research community and is managed by the NCI Merit Allocation Scheme. Researchers with a need for MASSIVE’s extensive rendering and visualization capabilities will be given priority, as will those whose applications leverage GPU acceleration. The next call for proposals starts in November for access in 2012, but early access may be sought by sending an email request to info@massive.org.au. Additional information regarding the allocation process is available at www.massive.org.au/access.

PRACE Now Accepting Applications for Supercomputing Time

The Partnership for Advanced Computing in Europe (PRACE), which provides Europe with access to cutting-edge supercomputing resources, is now accepting submissions for its third call for proposals. Successful applicants will be able to access a total of 3 Tier-0 supercomputers and 17 national Tier-1 systems.

This call marks the first time that PRACE affiliates will get to use the Tier-0 “HERMIT” supercomputer. This Cray XE6 system offers one petaflop peak performance and will be installed in the fall at the High Performance Center of University Stuttgart. A planned upgrade is already in the works for the 2013, which will supply “HERMIT” with an additional 3-4 petaflops of power, creating a system with a possible 5 petaflops of peak performance.

The one-petaflop IBM BlueGene/P system, JUGENE, based at Germany’s Jülich Supercomputing Centre, and the 1.6 petaflop Bull Bullx cluster, CURIE, hosted by the French research agency, CEA, will also be available as part of this allocation. And for the first time, seventeen Tier-1 systems are also being included in the PRACE call. These Tier-1 resources were previously overseen by DEISA (the Distributed European Infrastructure for Supercomputing Applications) and were part of DECI calls, which now fall under the purview of PRACE.

More information about the PRACE allocation process is available at www.prace-ri.eu/hpc-access. The current application period runs from May 2 – June 22, 2011.