DOE collaboration to accelerate high-performance computing

Nov. 8, 2004

DOE collaboration to accelerate high-performance computing

LIVERMORE, Calif. - Voltaire, Topspin and Intel have won an Advanced Simulation and Computing (ASC) PathForward award to develop a new open-source software stack for InfiniBand™, the U.S. Department of Energy's (DOE) National Nuclear Security Administration (NNSA) announced today. The software will comply with the InfiniBand (IB) specifications and accelerate the creation of the OpenIB Alliance software stack for High Performance Computing (HPC).

Under the terms of the award, Voltaire will lead a group of leading InfiniBand companies including Intel and Topspin Communications to accelerate the production of a single open-source software stack to meet the scalability, performance, portability, reliability and manageability requirements of large clusters for the NNSA national laboratories and the wider HPC community. The multi-year project will accelerate the efforts to include the OpenIB stack as part of the standard Linux kernel and is expected to include additional OpenIB contributors in later stages.

“We are delighted and honored to be selected by these prestigious computing institutions to lead a team to deliver the next generation InfiniBand software stack for the open source, HPC and enterprise communities,” said Ronnie Kenneth, CEO and chairman, Voltaire.

“The PathForward development effort, in conjunction with OpenIB, will enable customers and suppliers in the InfiniBand community to focus on a single common software stack that meets enterprise requirements as well as those for high-performance computing,” said Steve Poole of the Advanced Development team at Los Alamos National Laboratory, “and will enhance the prospects for wider market acceptance of the InfiniBand Architecture as the interconnect of choice. This project represents an outstanding example of the three DOE National Laboratories, Los Alamos, Lawrence Livermore and Sandia, partnering with industry to enhance global HPC.”

“OpenIB PathForward marks an important step for InfiniBand toward widespread adoption in High Performance Computing,” said Mark Seager, Principal Investigator for ASC platforms at Lawrence Livermore National Laboratory. “A true open-source solution supported by the DOE means the HPC, enterprise and storage industries can be assured of InfiniBand's robustness and interoperability across computing architectures and manufacturers. It also positions InfiniBand for deployment as a high-performance cluster interconnect for expanding the scope of science insight we will gain from our capacity computing.”

“The NNSA OpenIB PathForward will ensure that the OpenIB software stack meets the performance, scalability and portability requirements of the HPC community,” said Matt Leininger, staff scientist at Sandia National Laboratories and the lead ASC technical representative for this PathForward. “We look forward to working closely with the open source community to develop a high-performance InfiniBand software stack for Linux.”

The OpenIB Alliance will submit certain modules of this stack to, requesting their inclusion in the 2.6 release, and will request Linux distributors such as Red Hat, SUSE to include these modules.

The Open IB PathForward is a multi-year project with milestones that include reducing the latencies of kernel-level drivers, increasing the scalability and performance of the upper layer protocols for message passing, file I/O, and cluster interconnect topology management. Throughout the project the team will optimize routing for multi-layer, multi-topology, IB networks while improving redundancy, robustness and reliability. In addition, the project will utilize the OpenIB software repository, creating an open-source community compliant software structure, problem tracking, maintenance and documentation capabilities.

OpenIB Alliance
The OpenIB Alliance is an industry association chartered to deliver a single open-source Linux-based software stack for deploying InfiniBand. Founded in June 2004, the OpenIB Alliance is comprised of technology vendors and end user organizations including Dell; Engenio Information Technologies, Inc, IBM; InfiniCon; Intel; Lawrence Livermore National Lab; Mellanox; Network Appliance; Los Alamos National Laboratory, Sandia National Laboratories; Sun Microsystems, Inc.; Topspin; Veritas and Voltaire.

NNSA PathForward
PathForward is a technology vendor partnership project of the National Nuclear Security Administration's Advanced Simulation and Computing (ASC) Program - an integral and vital element of the nation's Stockpile Stewardship Program, by which the United States ensures confidence in the safety, performance and reliability of its nuclear stockpile. ASC provides the integrating simulation and modeling capabilities and technologies needed to combine new and old experimental data, past nuclear test data, and past design and engineering experience into a powerful tool for future design assessment and certification of nuclear weapons and their components. Through PathForward, ASC strives to make strategic, targeted investments with its vendor partners to accelerate the development of hardware and software technologies needed to ensure that complete, balanced systems for capability and capacity computing at Lawrence Livermore National Laboratory, Los Alamos National Laboratory and Sandia National Laboratories are available in the marketplace for out-year procurements by the program.

Founded in 1952, Lawrence Livermore National Laboratory is a nuclear security laboratory, with a mission to ensure national security and apply science and technology to the important issues of our time. Lawrence Livermore National Laboratory is managed by the University of California for the National Nuclear Security Administration/Department of Energy.