April 21, 2011
The Weekly Top Five features the five biggest HPC stories of the week, condensed for your reading pleasure. This week, we cover TACC's agreement with Intel to develop the company's MIC processor line; SGI's Japanese gigs; the High Performance Computing Center Stuttgart's new Cray systems; the mapping of red blood cells in the brain; and the creation of better fusion models using the Jaguar supercomputer.
TACC, Intel Collaborate for Open Science
Today the Texas Advanced Computing Center (TACC) at The University of Texas at Austin announced a new partnership with Intel to help the open science community prepare for Intel's upcoming "many integrated core" (MIC) processor line. While there are 100 other Intel partners working on software development for the MIC processors, TACC is the first National Science Foundation (NSF) TeraGrid institution to partner with Intel for the benefit of the open-science community.
According to the release, TACC has been provided with a software development platform for pre-production "Knights Ferry" MIC processors and is already porting applications. Later this year, TACC and Intel will build a Knights Ferry-based cluster to explore scalability issues. The partners will report on their efforts at the November 2011 Supercomputing Conference. In addition, TACC will have early access to the first commercial MIC processors, codenamed "Knights Corner," to ensure application performance.
Knight's Corner will employ Intel's 22-nm manufacturing process and will enable 50 cores on a single chip. While no official release date has been announced, there's talk the commercial chips will debut in the second half of 2012. The chips are being targeted at applications with a high degree of data parallelism. Examples include molecular dynamics and quantum chemistry, as well as data-intensive applications like seismic imaging, sensor network analysis, and real-time analytics.
HPCwire Editor Michael Feldman reveals additional details on the architecture:
MIC represents Intel's entry into the HPC processor accelerator sweepstakes, as the company attempts to perform an end-run around GPU computing. Mainly thanks to NVIDIA, over the last few years GPU computing, aka GPGPU, has become a mainstream HPC solution across workstations, clusters and supercomputers. They rely on specialized programming environments, like CUDA and OpenCL, to develop software on those platforms.
As suggested by its name, MIC is essentially an x86 processor, with more cores (but simpler ones) than a standard x86 CPU, an extra-wide SIMD unit for heavy duty vector math, and four-way SMT threading. As such, it's meant to speed up codes that can exploit much higher levels of parallelization than can be had on standard x86 parts.
Feldman spoke with TACC's deputy director Dan Stanzione, who was optimistic regarding the product's x86 compatibility. As Stanzione confided to Feldman, "Moving a code to MIC might involve sitting down and adding a couple of lines of directives that takes a few minutes. Moving a code to a GPU is a project."
SGI, Big in Japan
This week SGI announced it was selected by Japan's Semiconductor Energy Laboratory (SEL) to provide an SGI Altix ICE 8400 system for semiconductor research and development applications. The supercomputer will support new technologies like thin-film integrated circuits, liquid crystal and electroluminescent displays, semiconductor thin-film transistors, solar cells, and batteries.
Outfitted with 3,840 Intel Xeon 5600 series processors and up to 15 terabytes of memory, the new compute cluster will be about ten times faster than its predecessor. Officials say the system will be operational in July.
Semiconductor Energy Laboratory, which has worked with SGI Japan in the past, cited SGI's solid track-record and SGI Japan's reputation for technical support as factors in their decision-making process.
Last Friday, SGI was selected to work with Bull on a 1.3 petaflop HPC system being installed at the International Fusion Energy Research Center in Rokkasho, Japan.
The High Performance Computing Center Stuttgart Orders Two Petascale Crays
The High Performance Computing Center Stuttgart (HLRS) of the University of Stuttgart, part of the larger Gauss Centre for Supercomputing (GCS), will soon deploy a one-petaflop Cray XE6 supercomputer known as "Hermit." Since the Gauss Centre for Supercomputing is a PRACE member institution, the new Cray supercomputer will be available to researchers, scientists and engineers throughout Europe.
This one-petaflop Hermit system, which is scheduled to debut in the fall of this year, will be followed by a 4-5 petaflop supercomputer as part of the second half of the project, to take place in 2013. The future Cray architecture is code-named "Cascade."
HLRS Director Michael Resch commented on the center's partnership with the big-league supercomputer-maker:
"Cray is just the right partner as we enter the era of petaflops computing. Together with Cray's outstanding supercomputing technology, our center will be able to carry through the new initiative for engineering and industrial simulation. This is especially important as we work at the forefront of electric mobility and sustainable energy supply."
HLRS is a key member of the Gauss Centre for Supercomputing (GCS), an alliance of three major supercomputing centers in Germany that together represent one of the world's largest supercomputing resources. In addition to being one of the leading centers of the Partnership for Advanced Computing in Europe (PRACE) initiative, HLRS is the only major European HPC organization to work directly with industrial partners in automotive and aerospace engineering. These new installations will increase the overall capacity of the PRACE Research Infrastructure.
Researchers Create Detailed Blood Flow Models
A team of scientists from Brown University and the U.S. Department of Energy's (DOE) Argonne National Laboratory are using the lab's Blue Gene/P supercomputer to map the movement of red blood cells in the hopes that it will lead to better diagnoses and treatments for patients with blood flow disorders. The research was made possible through the DOE's Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program, which awarded 50 million processor-hours to the project.
With advances in supercomputing, researchers can now create detailed models of blood flow down to the molecular level, enabling doctors to better understand how heart and blood diseases can be treated. This field of study is known as "biophysics" since "the forces that govern red blood cells' movements at this level are best described by the laws of physics."
Joe Insley, team member and principal software developer at Argonne, comments on the achievement:
"Previous computer models haven't been able to accurately account for, say, the motion of the blood cells bending or buckling as they ricochet off the walls. This simulation is powerful enough to incorporate that extra level of detail."
Part of the study involves mapping the movement of red blood cells in the brain. The team used similar modeling last year to discover that the malaria parasite causes its victims' red blood cells to be 50 times stiffer than normal. According to the announcement, the research on blood flow in the brain could lead to treatments for diseases that affect blood flow, such as malaria, diabetes and HIV.
Another part of the study is looking at the relationship between cerebrospinal fluid and blood flow in the brain. When this system breaks down, it can put pressure on brain tissues, leaving the brain vulnerable to damage.
The researchers, led by G. E. Karniadakis, used Argonne's Blue Gene/P supercomputer, located at the Argonne Leadership Computing Facility (ALCF). The IBM machine is capable of performing 500 trillion calculations per second, enough power to solve the most challenging science problems.
Jaguar Supercomputer Heats Up Fusion Reactions
A team of researchers is using Oak Ridge National Laboratory's Jaguar supercomputer to study fusion reactions. The reactions produce helium from hydrogen and release energy in the process, and could be used to ignite ITER, an experimental fusion reactor under construction in southern France.
Zhihong Lin of the University of California-Irvine is working with General Atomics researcher Ron Waltz on the project. As part of the the Department of Energy's INCITE program, the team received three years of processor time on the Oak Ridge Leadership Computing Facility's Cray XT5 Jaguar, which can process two quadrillion calculations per second. These fusion simulations use between 5,000 and 50,000 of Jaguar's 224,256 processing cores.
The research sheds light on the role of turbulence in a fusion plasma. Turbulence can threaten the fusion reaction by allowing charged particles to cool. According to the announcement, "Lin's team is using simulations to develop ways of applying electromagnetic forces to overcome turbulence, heating the reactor, rather than cooling it."
The researchers have been working to create computer programs that will lead to more accurate and useful fusion plasma simulations. According to Lin, a complete model will be capable of "simultaneously simulating all turbulent interactions between the particles in a fusion reaction." His goal is to complete such a model by 2012.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?