OLCF staff attend the first session of the “Best Practices for HPC Software Developers” webinar series, jointly hosted by the OLCF, the ALCF, NERSC, and the IDEAS project, on May 4, 2016 at ORNL. Other participants joined the session from their respective locations via a teleconferencing service. Picture: (clockwise from the front) Ramakrishnan Kannan, Sherry Ray, Gustav Jansen, Nathan Rowe, Bruno Turcksin, Damien Lebrun-Grandie, Kristin Safi, Judy Hill, and David Bernholdt.
The IDEAS project, Argonne, NERSC, and Oak Ridge jointly offer best practices webinars for software development
The Oak Ridge and Argonne Leadership Computing Facilities (OLCF and ALCF), the National Energy Research Scientific Computing Center (NERSC), and the Interoperable Design of Extreme-scale Application Software (IDEAS) project are presenting a series of webinars—“Best Practices for HPC Software Developers”—to help users of high-performance computing (HPC) systems carry out their software development more productively.
The IDEAS project focuses on increasing software sustainability and developer productivity for high-performance applications in computational science. IDEAS is a collaboration among researchers at seven US Department of Energy (DOE) laboratories—Argonne National Laboratory (ANL), Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory (LLNL), Los Alamos National Laboratory (LANL), Oak Ridge National Laboratory (ORNL), Pacific Northwest National Laboratory (PNNL), and Sandia National Laboratories (SNL)—and the Colorado School of Mines.
One challenge IDEAS researchers are addressing is the development of an extreme-scale scientific software development kit (xSDK). The xSDK brings together a number of widely used US Department of Energy (DOE) numerical libraries into a common configuration and build environment to simplify usage and interoperability for software developers who rely on these libraries.
Another element is methodologies for software productivity, where researchers adapt, develop, and deploy best practices—drawing on approaches in the software engineering community and translating them into resources that are useful for the HPC computational science community. All work in the project is motivated and validated by applications in extreme-scale science, with primary emphasis on use cases in terrestrial modeling. Another key aspect of work is outreach, which involves opening a dialogue with the community in the form of training and other events. Our outreach efforts draw upon the experience and content developed in the IDEAS project methodologies, xSDK, and use-case collaborations.
The webinar series is the IDEAS project’s first broad outreach and training program. “This activity will allow people with substantial expertise in HPC to share their experiences with the community and engage in a conversation,” said David Bernholdt, ORNL Computer Science Research Group leader and IDEAS project outreach lead. “We’re trying to reach out, provide specific suggestions to help developers, and raise awareness that the topic of software developer productivity is an essential aspect of overall scientific productivity.”
Judy Hill, the OLCF liaison for the IDEAS project, added, “Much of the software that runs on Titan and at other facilities is for scientific computing, and we haven’t really focused our training efforts on the sustainability of that software or the engineering process. Historically, we haven’t done a lot to offer basic tools and processes that researchers could use to make their software development easier, which in turn would make it easier for facilities to work with them. That’s a niche that this project is filling by bringing attention to this cause. It’s beneficial not only to researchers and developers but also to the computing facilities.”
For the first session on May 4, Argonne computer scientist Anshu Dubey provided an overview of the project’s main tenets.
“We had a great response to the first webinar, with more than 150 people across the world participating,” said Katherine Riley, ALCF Director of Science and host of the May 4 session. “The strong interest in this series illustrates that there is a growing awareness of the importance of software design and engineering methodologies in the computational science community.”
In the second of the seven planned sessions on May 18, Argonne distinguished fellow Barry Smith covered developing, configuring, building, and deploying HPC software. Four of the presenters in the webinar series come from the IDEAS project. The other three come from the three DOE Office of Science User Facilities focused on supercomputing.
“NERSC is delighted to collaborate with the IDEAS project, ALCF, and the OLCF to present such a high-quality series of talks,” said Rebecca Hartman-Baker, the NERSC liaison for the IDEAS project. “The speakers lined up for this series are all first-rate and have very interesting ideas as well as practical advice to share with the HPC community.”
The series is being presented through a teleconferencing service to encourage participation from the audience. Attendees have opportunities to ask questions, and several sessions will include interactive demonstrations. The sessions are being recorded and posted online. So far, the webinar series has had a tremendous response with more than 300 registrations. For more information and to register, visit https://www.olcf.ornl.gov/training-event/webinar-series-best-practices-for-hpc-software-developers/.
“DOE has accomplished a tremendous amount through HPC, and the role of computational science is increasing rapidly,” Bernholdt said. “But as the expectations for computational science get higher, and the computer systems become more complex, the task of developing robust, high-quality software becomes more difficult. It becomes even more challenging as we look to future machines with greater architectural diversity, including the OLCF’s Summit, ALCF’s Aurora, and NERSC’s Cori, and further to exascale systems. It’s important to focus on the software development process and exchange information so that we can develop better software, since making software more reliable and scientifically robust will help us produce better science more quickly.” – Miki Nolin
Oak Ridge National Laboratory is supported by the US Department of Energy’s Office of Science. The single largest supporter of basic research in the physical sciences in the United States, the Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit science.energy.gov.