HPC peers share experience and vision for computing in atmospheric science

By Anke Kamrath
09/30/2015 - 12:00am

The 13th biennial International Computing for the Atmospheric Sciences Symposium (iCAS) was held in Annecy, France on 13−17 September 2015. This year’s program included a wide range of interesting topics and excellent speakers. Now available at the iCAS2015 website are the abstracts and slides from the presentations (follow the “Program” link at left). There were about 60 participants from nine countries spanning four continents, with balanced representation and presentations from High Performance Computing (HPC) leaders, technology vendors, and scientists who rely on supercomputing facilities for their research. One speaker, Bill Kramer of the National Center for Supercomputing Applications (NCSA), captured the character of the symposium when he stated, “This symposium is always good because it takes a high-level view across the computing and application landscape.”

Two great retrospective presentations described changes over the last few decades in both climate modeling (“Climate Change Modeling: Present and Future” by Warren Washington, NCAR) and technical computing (“40 Year Perspective of Technical Computing at IBM” by Don Grice, IBM). These presentations were complemented by an excellent future-looking presentation titled “Cloud Resolving Models or Futures in Climate Modeling (aka ‘The Grey Zone’)” by David Randall, Colorado State University.

Many talks focused on solutions for the “Big Data” challenges that include large data volumes, workflow bottlenecks, model performance, data download, and analysis capabilities. These challenges arise from analysis and workflows in large multinational efforts such as the past Coupled Model Intercomparison Project 5 (CMIP5) and the future CMIP6, but also from ever-increasing model resolutions in numerical weather prediction, Earth System process studies, and global reanalysis projects. Most organizations attending iCAS2015 are developing novel parallel I/O and coupled HPC strategies, along with dedicated analysis platforms that are supported by high performance infrastructure to tackle the data challenges.

Several presentations shared information on facilities, computer procurements, and resources: the National Energy Research Scientific Computing Center (NERSC) CORAL System and Implications for Future HPC Hardware and Data Centers, UK Academic Infrastructure, System Co-design at MeteoSwiss, Canadian Meteorological Centre HPC Renewal Initiative, the European Center for Medium-range Weather Forecasting (ECMWF) System Update, UK Met Office HPC, The German Climate Computing Centre (Deutsches Klimarechenzentrum), DKRZ System, and Lessons Learned at the NCAR Wyoming Supercomputing Center facility (NWSC).

Two presentations by leading visualization experts gave us an impressive and beautiful view into a detailed visualization of a storm (“Simulating and Visualizing Tornadic Thunderstorms at Ultra High-Resolution” by Leigh Orf, University of Wisconsin) and also a look at future (exascale) visualization challenges (“Exascale Visualization: Get Ready for a Whole New World” by Hank Childs, University of Oregon). Attendees agreed that visualization is having an increasing impact on the HPC facilities and is an essential tool for understanding the data being generated.

The symposium, as always, was topped off with excellent networking and knowledge sharing in a venue that fosters great interactions and conversations with colleagues in historic Annecy, France. We hope to see all the attendees back and welcome new participants at the next iCAS in September 2017.

EF5 airflow
You’re looking at unsteady particles (in yellow) dropped every two seconds into the forward flank of a very strong convective storm right along the leading edge of the cold pool, combined with vorticity magnitude (the volume-rendered surface). The EF5 tornado (strongest tornado category) and other subvortices, some of which are oriented horizontally, are shown in green. CISL’s VAPOR 3D visualization software package was used to produce this image. (Image courtesy of Leigh Orf, University of Wisconsin.)