Skip to end of metadata
Go to start of metadata

ECSS Work Plan

Title:  Weather Simulation on Petascale Computational Resources

PI:  John Q Public

PI Institution:  The University of the Americas

ECSS Consultant:  Jane Doe, SDSC; John Smith, PSC

Allocation Start/End dates:  Jan 1, 2009 – Dec. 31, 2009

Abstract:  Weather is an unavoidable fact of everyday life.  It informs a number of important decisions such as when to schedule an outdoor event, plant crops, or even what to wear.  However, the most important events to anticipate are extreme events such as floods, droughts, hurricanes, or tornados.  These events have large, long-lasting effects which could be mitigated if ample preparation time is available.    Weather simulations over both short and long time scales are necessary to adequately predict the likelihood of these events and guide preparedness/warning programs.  The use of petascale computational resources will enable both real-time weather forecasts as well as long-term climate statistics to produce the hazard maps and warnings. 

Objective:  Optimize the WFC (Weather Forecasting Code) to produce real-time forecasts (computation time << forecast interval).  Additionally, optimize the WFC to provide more reliable hazard maps utilizing more computational resources. 

Scope:   The WFC requires a short computational time to produce real-time forecasts which are useful once complete.  The major obstacle with respect to this goal is the time required to migrate data to the computational resource in preparation for computation.  A possible solution is an automated system to migrate and preprocess data in anticipation of the computation.  Additionally, a parallel I/O solution within WFC may greatly decrease the amount of time required for computation.  Finally, more efficient data decomposition or task oriented parallelism may improve WFC’s computational efficiency. 

Sustainability:  WFC is a freely available software package which has been significantly augmented within the project group.  This development version is kept under revision control and documented within our research group.  All code modifications and enhancements will be applied to the software repository after validation.


  1.       An automated framework to migrate and preprocess input data. – ESRT
  2.       Implement parallel I/O solution for file I/O. – ESCC
  3.       Improve computational efficiency via investigation of data decomposition and task oriented parallelism. – ESRT


  1.       Development of an automated data migration and preprocessing framework. – Deliverable 1
    1.       Success:  Identification of a data migration solution and required functionality of preprocessing utility.
    2.       Effort:  Moderate data collection.  Investigation and collection of framework requirements.
    3.       Estimated completion:  March 2009 (1 Quarter)
  2.       Development of an automated data migration and preprocessing framework. – Deliverable 1
    1.       Success:  A functional, automated framework which migrates and preprocesses data prior to computation.
    2.       Effort:  Major development required.  Development of preprocessing utility and identification of data migration solution is required.
    3.       Estimated completion:  May 2009 (2 Quarter)
  3.       Implement parallel I/O solution for preprocessed data. – Deliverable 2
    1.       Success:  A parallel I/O solution is implemented in WFC for the preprocessed data which reduces the total computational time by at least half.
    2.       Effort:  Moderate development required.  The identification of a suitable file format is required.
    3.       Estimated completion:  August 2009 (3 Quarter)
  4.       Optimize WFC’s data decomposition paradigm to improve performance and scalability. – Deliverable 3
    1.       Success:  Identify and implement better performing and/or more scalable data decomposition paradigms in WFC.  Improve performance (including I/O savings) by  a factor of three and allow the utilization of at least 500 processes.
    2.       Effort:  Moderate investigation and implementation required.
    3.       Estimated completion:  December 2009 (4 Quarter)


  1.       ECSS Staff – Jane Doe, SDSC (15% effort); John Smith, PSC (15% effort)
    1.      Data migration consultant – Jane Doe, SDSC (ESRT)
    2.      Parallel I/O  – John Smith, PSC (ESCC)
    3.      WFC profiling – John Smith, PSC (ESRT)
    4.      WFC optimization – Jane Doe, SDSC (ESRT)
  2.       Project Group – A. Smith (6 months); C. Doe (6 months)
    1.      Data migration and preprocessing framework – A. Smith
    2.     Parallel I/O implementation – A. Smith
    3.     WFC profiling and optimization – C. Doe
  3.       Risks and Mitigation
    1.     Parallel I/O implementation is highly correlated with data decomposition paradigm.
      1.    Mitigation:  Milestone 3 will be completed before implementation of parallel I/O.
    2.     Development of data migration framework requires substantial effort
      1.   Mitigation:  Parallel I/O implementation will be completed as last milestone.  Milestone 3 will be eliminated. 
  • No labels