CISL 2014 Annual Report released

By Brian Bevirt
01/09/2015 - 12:00am

NCAR’s Computational and Information Systems Laboratory (CISL) structures its planning efforts to provide world-class service, science, and education for the Earth System sciences research community. CISL also tracks the ways it produces broader impacts by diversifying the research workforce and providing leadership in cyberinfrastructure and high performance computing. CISL’s 2014 Annual Report highlights its most recent contributions and successes in all of these areas.

Exploring data reduction strategies that are suitable for scientific data sets is a key component of CISL’s Big Data research agenda. These two visualizations of a synthetic data set that has been compressed 64:1 show a qualitative comparison between output from a widely used compression scheme (left) with results from a new encoder (right). The image on the left clearly exhibits data compression artifacts, while the one on the right is visually indistinguishable from imagery generated from the original data.

For FY2014, CISL focused on the ways it is anticipating and confronting the “Big Data” challenges of the Earth System sciences. This year’s report describes numerous changes and refinements CISL has made to address the challenges of ever-larger data sets:

  • Changed and upgraded hardware and software for Yellowstone, Caldera, Geyser, GLADE, and HPSS to improve performance and add new capabilities.
  • Upgraded the entire Yellowstone environment’s full system software stack.
  • Refined the data-centric workflows through NCAR’s supercomputing equipment.
  • Enhanced our observational and model-output research datasets.
  • Advanced high-speed sharing of large data sets through multiple science gateways and by implementing Globus Plus software.
  • Upgraded local, regional, and national network interfaces to increase performance and reliability and to include more universities.
  • Improved data analysis and visualization tools such as NCL and VAPOR.
  • Developed parallel data processing tools such as PyReshaper and LatticeKrig to improve model output handling.
  • Added new software to make the Data Assimilation Research Testbed more efficient.
  • Continued increasing the effectiveness of data-compression methods.
  • Added new data service capabilities that allow NARCCAP users to access only the data they need from very large datasets.
  • Extended statistical methods to analyze large spatial datasets.

These services, tools, and methods combine to support our communities in extracting scientific knowledge from the petabytes of data provided by NCAR. End-to-end workflows employing these parallel tools, services, and methods help scientists produce results more quickly for broader audiences.

The enormous datasets used to produce this image were visualized on CISL’s Geyser cluster using VAPOR software. It is conventionally understood that winds in the atmosphere generate ocean waves such as the large waves produced by hurricanes. The forces that the ocean waves exert on the atmosphere, however, are more subtle and not as well understood. The simulation visualized in this image required 12 million compute hours on Yellowstone to help researchers understand how the effects of ocean waves ripple up into the atmospheric boundary layer.

The FY2014 CISL Annual Report also describes a variety of other highlights from the past year. CISL launched several new forward-looking initiatives to prepare for our future challenges:

  • Started the procurement process for the supercomputer that will succeed Yellowstone in 2017.
  • Deployed the HPC Futures Lab to extend CISL’s research in high performance computing, which will improve the current computing environment and help us assess technology that may be used in future systems.
  • Launched the Strategic Parallel and Optimization Computing initiative to increase the performance and efficiency of NCAR’s community codes – CESM, WRF, and MPAS – on Yellowstone, and to prepare these codes for future architectures.
  • Developed a simple way for researchers to transfer and share large data sets via national and international networks using Globus Plus software.
  • Formed an Intel Parallel Computing Center focused on Weather and Climate Simulation (IPCC-WACS) in collaboration with the University of Colorado at Boulder to increase the performance of atmospheric applications using advanced microprocessor technologies, and to help train the next generation of scientists and engineers who will apply these new technologies.

CISL continues to offer progressive and well-targeted education programs to enrich and expand our research community. SIParCS, TOY, and RSVP are mature programs that encourage young scientists and engineers to plan for and succeed at careers in the computational and Earth System sciences. CISL now offers a selection of webinars and online training for HPC, NCL, and VAPOR users. The NWSC visitor center continues to attract educational groups and the general public. In FY2014 it hosted 47 large groups and a total of 1,713 visitors, and it has become a preferred venue for regional STEM events.

CISL’s education programs integrate research and education and teach the technical skills that students and faculty need to make effective use of advanced cyberinfrastructure. These programs also promote diversity, enhance CISL’s culture of teaching and mentorship, and stimulate collaborations with the university community. During one of IMAGe’s Theme of the Year events at NCAR’s Mesa Lab, Michael Angus (left), a graduate student at North Carolina State University presents his research poster to Kevin Trenberth, a Distinguished Senior Scientist at NCAR. More than 90 people participated in this June 2014 Fourth Workshop on Understanding Climate Change from Data sponsored by NSF Expeditions in Computing.

As part of its diversity program, CISL conducts mission-appropriate outreach to integrate education and research, broaden participation in Earth System sciences, and develop the future STEM workforce. CISL does outreach through site visits, conferences, job fairs, and presentations to tribal colleges and minority-serving institutions. In FY2014, CISL enhanced interactions with its existing network of contacts and developed connections with new institutions and conferences. These networks connect the work of NCAR scientists and engineers with interested students in minority groups, EPSCOR states, and two-year colleges. This year, CISL launched its first summer externship program to teach parallel computing concepts to nontraditional students who have obligations that do not allow them to spend an entire summer in Boulder for an internship.

CISL provides far more than balanced, user-friendly computational and data environments designed for the evolving requirements of the Earth System sciences. CISL also develops and delivers high-quality science and education programs to help secure the future of our scientific enterprise and the communities we serve.