Back to Table of contents

Primeur weekly 2016-11-21

Exascale supercomputing

Brookhaven Lab to lead and partner on DOE exascale computing projects ...

Quantum computing

World's fastest quantum simulator operating at the atomic level ...

Tracking the flow of quantum information ...

Tyndall technology lights the way for quantum computing ...

New attacks on location-based quantum cryptography ...

Focus on Europe

Fall 2016 edition of the e-IRG newsletter available ...

Fujitsu in joint intercontinental data centre coordination field trial using cold data storage ...

Asetek announces first installation using InRackCDU technology ...

Middleware

Altair PBS Cloud is Altair's latest technology to further HPC appliance solutions ...

Intel HPC Orchestrator to include PBS Pro for workload management ...

Altair expands the availability of PBS Professional ...

Adaptive Computing launches new Reporting & Analytics solution ...

Adaptive Computing opens up platform of products ...

Adaptive Computing improves ease-of-use & cost efficiency in Moab HPC Suite 9.1 release ...

Hardware

PGI accelerator compilers for POWER architecture enable easy on-ramp to GPU acceleration with POWER8 and NIVIDA NVLink ...

Liquid silicon: Computer chips could bridge the gap between computation and storage ...

Cray expands customer base for Cray Urika-GX system and previews upcoming software release ...

Fujitsu develops analysis technology to improve communication performance of virtual networks ...

Applications

NVIDIA teams with National Cancer Institute, U.S. Department of Energy to create AI platform for accelerating cancer research ...

Supercomputer simulations help develop new approach to fight antibiotic resistance ...

Earlham Institute receives supercomputing award for wheat research ...

Chinese research team that employs high performance computing to understand weather patterns wins 2016 ACM Gordon Bell Prize ...

Cray systems power deep learning in supercomputing at scale ...

IBM and Topcoder bring Watson to more than one million developers ...

IBM Watson Health and Broad Institute launch major research initiative to study why cancers become drug resistant ...

Julia for Astronomy: Parallel computing with Julia on NERSC supercomputer increases speed of image analysis 225x ...

Julia Computing and IBM present Julia for Deep Learning at SC16 ...

TOP500

Jefferson Lab's newest cluster makes TOP500 list of fastest supercomputers ...

Hewlett Packard Enterprise/SGI Cheyenne and Pleiades systems named on TOP500 list ...

The Cloud

Fujitsu and SUSE expand strategic alliance to develop and support open source products ...

IBM closes acquisition of Sanovi Technologies ...

Tencent Cloud joins IBM and Mellanox to break data sorting world records ...

Brookhaven Lab to lead and partner on DOE exascale computing projects


Clockwise from left: Wei Xu, Klaus Mueller, Shinjae Yoo, and Kerstin Kleese van Dam of Brookhaven Lab's Computational Science Initiative are partnering with Argonne and Oak Ridge National Laboratories and several universities on CODAR, one of four co-design centre projects that the U.S. Department of Energy's Exascale Computing Project (ECP) recently funded. Not pictured: Brookhaven team members Barbara Chapman, who is also leading an ECP-funded software development project called SOLLVE, and Abid Malik, both of the Computational Science Initiative. Credit: Brookhaven National Laboratory.
16 Nov 2016 Upton - The U.S. Department of Energy's (DOE) Exascale Computing Project (ECP) just awarded $34 million in first-year funding to 35 software development proposals and $12 million to four co-design centre proposals. Scientists at the DOE's Brookhaven National Laboratory are leading one of these software development projects, SOLLVE, and contributing to one of these co-design centres, CODAR. SOLLVE will focus on OpenMP functionality for exascale computing; CODAR will concentrate on online data analysis and reduction at the exascale.

Exascale computing refers to computing systems that are at least 50 times faster than the USA's most powerful supercomputers used today. The ECP is responsible for the planning, execution, and delivery of technologies - including software, applications, hardware, and early testbed platforms - needed for the U.S. to effectively design and run the exascale systems of the future.

"The funding of these software development projects and co-design centres, following our recent announcement for application development awards, signals the momentum and direction of ECP as we bring together the necessary ecosystem and infrastructure to drive the nation's exascale imperative", stated ECP Director Paul Messina.

In September 2016, ECP announced the selection of 15 fully funded and seven seed-funded application development projects, with a total funding of $39.8 million. Brookhaven Lab is a partner on two of the fully funded projects: one in computational chemistry, and the other in computational physics.

"Brookhaven is proud to contribute at all levels to this exciting capability development that will enable DOE to meet its scientific and national security mission needs over the next decade", stated Kerstin Kleese van Dam, director of Brookhaven's Computational Science Initiative, which is addressing the challenges of exascale computing as part of its research portfolio.

Most ECP application development projects plan to use OpenMP (Open Multiprocessing), an industry standard for programming shared-memory parallel computers, as part of their strategy for reaching exascale performance levels. However, with respect to the requirements of these applications, there are critical gaps in OpenMP functionality that have resulted from the extremely rapid changes in hardware technology and programming languages relevant to exascale computing.

Here is where the Brookhaven-led software development project called SOLLVE - for Scaling OpenMP via LLVM (Low-Level Virtual Machine) for Exascale Performance and Portability - comes in.

"Obtaining exascale performance on tomorrow's largest computer systems will only be possible if the resources in each compute node are well utilized. OpenMP is key to achieving node resource efficiency on today's computers. In this development project, we intend to get OpenMP ready for the challenges of tomorrow", stated project lead Barbara Chapman, director of Brookhaven's Computer Science and Mathematics Group who is also a professor in the Applied Mathematics and Statistics Department and the Computer Science Department at Stony Brook University. Barbara Chapman's team includes collaborators from Argonne, Lawrence Livermore, and Oak Ridge National Laboratories, as well as Rice University and the University of Illinois at Urbana-Champaign.

The SOLLVE team will design, implement, and help standardize key OpenMP functionality features identified by ECP application developers. These features include the ability to program new kinds of memory, efficiently move complex data structures between different memories, create portable code that maintains high performance across various exascale architectures, and support the latest standards in programming languages - especially C++, which ECP applications are trending toward. The team will also ensure that OpenMP's full feature set is available to ECP participants.

Designing an exascale computer system involves the integration of many components: hardware, software, numerical methods, programming models, algorithms, and applications. A coordinated development process in which developers co-design these components within the context of constraints such as development costs, power efficiency, and scalability is critical to reaching exascale performance. "Co-design centres" help guide this process.

The co-design centre that Brookhaven is partnering on - Co-Design Centre for Online Data Analysis and Reduction at the Exascale, or CODAR - is focused on developing and implementing services to analyze and reduce data online before the data are written to disk for possible further offline analysis.

Exascale systems are projected to provide unprecedented increases in computational speed, but the input/output rates of transferring the computed results to storage disks are not expected to keep pace. Given this disparity, it will be infeasible at the exascale for scientists to save all of their scientific results for offline analysis.

With its partners - Argonne (project lead) and Oak Ridge National Laboratories and Brown, Rutgers, and Stony Brook universities - Brookhaven will develop the services required to output just the data needed by the application to enable scientific insight at the exascale. CODAR team members, many of whom are partners on ECP application and software development projects, will develop data analysis and reduction methods tailored to the needs of these projects.

"The Brookhaven team will focus its efforts on developing machine learning solutions for data reduction and analysis, creating visual analytics tools to assess and steer the reduction and analysis process, and enabling the reproducibility of this process so that the integrity of scientific results can be maintained. Underpinning these efforts will be the development of runtime system support for the data-intensive applications", stated Kleese van Dam, who is leading Brookhaven's efforts on the project. Barbara Chapman, Abid Malik, Klaus Mueller, Wei Xu, Shinjae Yoo are on the Brookhaven team.

The ECP is a collaborative effort of two DOE organisations - the Office of Science and the National Nuclear Security Administration. As part of President Obama's National Strategic Computing Initiative, ECP was established to develop a capable exascale ecosystem, encompassing applications, system software, hardware technologies and architectures, and workforce development to meet the scientific and national security mission needs of DOE in the mid-2020s timeframe.

Brookhaven National Laboratory is supported by the Office of Science of the U.S. Department of Energy. The Office of Science is the single largest supporter of basic research in the physical sciences in the United States, and is working to address some of the most pressing challenges of our time.

Source: DOE/Brookhaven National Laboratory

Back to Table of contents

Primeur weekly 2016-11-21

Exascale supercomputing

Brookhaven Lab to lead and partner on DOE exascale computing projects ...

Quantum computing

World's fastest quantum simulator operating at the atomic level ...

Tracking the flow of quantum information ...

Tyndall technology lights the way for quantum computing ...

New attacks on location-based quantum cryptography ...

Focus on Europe

Fall 2016 edition of the e-IRG newsletter available ...

Fujitsu in joint intercontinental data centre coordination field trial using cold data storage ...

Asetek announces first installation using InRackCDU technology ...

Middleware

Altair PBS Cloud is Altair's latest technology to further HPC appliance solutions ...

Intel HPC Orchestrator to include PBS Pro for workload management ...

Altair expands the availability of PBS Professional ...

Adaptive Computing launches new Reporting & Analytics solution ...

Adaptive Computing opens up platform of products ...

Adaptive Computing improves ease-of-use & cost efficiency in Moab HPC Suite 9.1 release ...

Hardware

PGI accelerator compilers for POWER architecture enable easy on-ramp to GPU acceleration with POWER8 and NIVIDA NVLink ...

Liquid silicon: Computer chips could bridge the gap between computation and storage ...

Cray expands customer base for Cray Urika-GX system and previews upcoming software release ...

Fujitsu develops analysis technology to improve communication performance of virtual networks ...

Applications

NVIDIA teams with National Cancer Institute, U.S. Department of Energy to create AI platform for accelerating cancer research ...

Supercomputer simulations help develop new approach to fight antibiotic resistance ...

Earlham Institute receives supercomputing award for wheat research ...

Chinese research team that employs high performance computing to understand weather patterns wins 2016 ACM Gordon Bell Prize ...

Cray systems power deep learning in supercomputing at scale ...

IBM and Topcoder bring Watson to more than one million developers ...

IBM Watson Health and Broad Institute launch major research initiative to study why cancers become drug resistant ...

Julia for Astronomy: Parallel computing with Julia on NERSC supercomputer increases speed of image analysis 225x ...

Julia Computing and IBM present Julia for Deep Learning at SC16 ...

TOP500

Jefferson Lab's newest cluster makes TOP500 list of fastest supercomputers ...

Hewlett Packard Enterprise/SGI Cheyenne and Pleiades systems named on TOP500 list ...

The Cloud

Fujitsu and SUSE expand strategic alliance to develop and support open source products ...

IBM closes acquisition of Sanovi Technologies ...

Tencent Cloud joins IBM and Mellanox to break data sorting world records ...