CISL 2007 annual report banner

Continuation of cyberinfrastruture reports from FY2006



Workshop on High Performance Computing for Geosciences Research

For several years, NCAR has been contributing to efforts to define the future requirements for both the science and the cyberinfrastructure of the geosciences. The findings of the High Performance Computing in the Geosciences workshop are published in the report issued in FY2007. Its recommendations included:

  • In general, there was consensus across the various disciplines that the geosciences community as a whole has problems suitable for petascale computing resources and that a need exists for access to large-scale computing resources.

  • Many of the discipline areas cited a need for the collaboratory to provide access to a spectrum of small- to large-scale computing facilities to handle a broad range of computations.

  • A number of the disciplinary breakout groups emphasized the need for both end-to-end community models and community data sets to be made available within the geosciences collaboratory.

  • The need for the collaboratory to provide end-to-end HPC research support was emphasized by the majority of workshop participants.

 

CISL's work on this effort aligns with NCAR's strategic goal to "Provide robust, accessible, and innovative information services and tools." This workshop series was based on work supported by the National Science Foundation under grant number 0631272.



Workshop on Geoscience Application Requirements for Petascale Architectures

The second and final workshop on Geoscience Applications Requirements for Petascale Architectures was held in San Diego, California on February 21-22, 2007 at the San Diego Supercomputing Center in La Jolla, California. This NSF-funded workshop brought together 35 computer scientists, application experts, and vendors to discuss a broad set of problems related to tackling petascale geoscience problems on the NSF's Track-1 and Track-2 systems.

The workshop series was centered around these broad questions:

  1. What are examples of important questions and conceptual challenges in the geosciences that illustrate the potential impact of access to a petascale computational facility?

  2. What strategies will ensure that the geoscience community is in position to take advantage of petascale computational capabilities?

  3. What resources are needed to get teams of geoscientists and computer scientists working together on petascale applications development, with the goal of having operational packages ready by 2011 when petascale resources will come on line?

 

The workshop series resulted in a final report with two general recommendations:

  1. A portfolio of candidate petascale applications should be established, and development funding should be provided for collaborative teams of geoscientists and computer scientists to prepare calculations that can both advance scientific discovery and run at petascale.

    Specifically, the following five application areas were identified at the workshop as candidates for that petascale applications portfolio:

    • Mantle dynamics (for example earthquake ground motion prediction using PetaShake at 25 meter, 2 Hz)

    • Climate (for example hemispheric "nature" runs to study energy spectrum in the atmosphere at the k-3 to k-5/3 kinetic energy spectral transitions)

    • Coupled ocean and weather (for example a hurricane eyewall calculation with turbulence and mixing at sub-10-meter resolution)

    • Space weather (for example a full-hemisphere coronal model to resolve and improve understanding of loop structure in a constrained coronal heating with loop structures model)

    • Ecological component of earth system modeling (for example adding plant cover to climate models)

     

  2. Suitable selection criteria should be applied to choose which specific calculations and teams from the above areas to invest effort and money into for the purpose of enabling viable petascale applications (via several years of effort starting as soon as possible).

    The following selection criteria were proposed in the workshop report:

    • Science impact

    • Strength and interdisciplinary nature of proposed team

    • Demonstrated plausibility for petascale

    • Needed investment in algorithms and model development

    • Risk of failure

 

CISL's work on this effort aligns with NCAR's strategic goal to "Provide robust, accessible, and innovative information services and tools." Funding was provided by NSF-ATM award 0540688 received through subcontract 2005-006559-01 with the University of Illinois.



Data center expansion advanced conceptual design

Two further revisions to the conceptual design were completed in FY2007. A warehouse facility located in Longmont, Colorado was further analyzed, and the base concept design was modified to fit this existing facility. The total cost of modifying an existing facility was found to have so many constraints that it was likely to cost more than constructing a new facility. In addition, a study was completed to compare the original site identified in the conceptual design with sites at the University of Colorado and at the Front Range Business Park in Cheyenne, Wyoming. The study found all three sites to be suitable from a technical perspective. The Wyoming site offered some additional flexibility.

The conceptual studies are complete, and NCAR is preparing to start detailed construction design work when approval is received from NSF. The conceptual design work was made possible through NSF Core funding.



Chilled water system upgrade

The chilled water system installation and testing have been completed, and it is now serving the blueice supercomputer in production usage. Water-cooled supercomputers are becoming necessary because of rapidly advancing transistor densities and processor speeds. Controlling ongoing increases in supercomputer heat output continues to challenge the chilled water system at NCAR's Mesa Lab facility.

This facility upgrade was made possible through NSF Core funding.