July 04, 2013
On this Independence Day in the United States we bring news of a new system set to boost the defense capabilities of the country via a new system housed at one of the nation’s top Air Force facilities.
The Air Force Research Laboratory Supercomputing Resource Center (DSRC) has a new addition to its fleet of supers via its new SGI ICE X system called Spirit, which will be housed at the Wright-Patterson Air Force base in Dayton, Ohio.
The top 20-level system, which is capable of 1.4 petaflops, will support various research, development, testing and evaluation projects, particularly on the aircraft and ship design fronts.
Spirit boasts 4,608 nodes and 73,728 Xeon cores humming at 2.6 GHz, as well as 146 TB of memory and 4.6 PB of disk space.
"By providing the technology solutions behind Spirit , SGI is further powering the work of the armed forces in the success of their missions and the safety of the men and women," said Jorge Titinger, president and CEO of SGI. "It is with great honor to see the results of the SGI ICE X system Spirit being included in the TOP500, further validating our work together and the effective supercomputing architecture we built to address complex HPC needs."
The DSRC is one of five HPC sites in the Department of Defense’s High Performance Computing Modernization Program. The goal of the program, which is managed on behalf of the DoD by the U.S. Army Engineer Research and Development Center, is to provide the DoD with supercomputing capabilities and support.
To learn about the new system we talked with Jeff Graham, Director, AFRL DRSC and John West, who heads up the HPC Modernization Program.
What is the current HPC environment look like at Wright-Patt and other bases?
Graham: The AFRL DSRC provides full spectrum services to customers across the DoD. We recently christened the modern Information Technology Complex which will house all of our future large scale systems - the first of which is the SGI ICE X system, Spirit. his new complex will give us eight MWs of battery/rotary Uninterruptible Power Supply coupled with backup diesel generators in order to insure consistent delivery of service and minimize the potential for damaging systems due to a power loss.
All of our human resources and monitoring activities will continue to take place across the street where our Mass Storage, Test and Development, and training resources are housed. In addition to HPCMP systems, we support several customer-funded, smaller clusters for specialized requirements in order to support critical research from the desktop, and mid-range clusters.
Can you comment on the decision to go with an SGI ICE X system over other alternatives?
West: The DoD HPCMP has a rigorous and well-refined process for its acquisition of new supercomputers that takes into account the most recent data on DoD mission requirements and user behavior in constructing an HPC system purchase portfolio which, in total, best meets the needs of the Department.
The purchase and installation of Spirit at the AFRL DSRC is part of the most recent acquisition completed by the HPCMP, which placed large-scale computing systems from IBM, Cray, and SGI at four supercomputing centers.
Each of the new systems acquired has unique attributes that match it most effectively to a portion of our user community's workload -- when managed as a portfolio of capability, they represent the most effective solution to meeting the computational requirements of the DoD.
Modeling and simulation are likely a core use of the machine--did you consider the applicability of acceleration/coprocessors on the system?
West: Not all modeling and simulation codes benefit from accelerators, due to the round trip latency that must be overcome between the processor and co-processor and the mismatch in the memory use patterns of many well-tuned HPC codes as compared to the memory structure of a co-processor.
The HPCMP is examining the applicability of specific accelerators to key applications and may make strategic investments based on these results in upcoming acquisitions. The HPCMP currently does have small platforms which contain some of the current commercially available options, including Telsa K20s and Xeon Phis, allowing side-by-side comparisons in order to determine the sensitivity of applications to different co-processor memory bandwidths (with and without ECC), memory sizes, hardware structures, and programming approaches.
Describe the HPCMP's role with this system--
West: As described in the answer to question 2, the AFRL DSRC is funded by the HPCMP and managed on our behalf by AFRL. The AFRL DSRC is one of the five DoD Supercomputing Resources Centers funded and managed by the HPCMP for the DoD's technical computing user community, and this acquisition is part of the HPCMP's annual supercomputing procurement. The acquisition is managed by the HPCMP program office with significant input from the DSRCs and representatives of our user community; funding for the acquisition is provided by the HPCMP program.
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?