Log on:
Powered by Elgg

Marlon Bright :: Blog :: PIRE Project Proposal

March 31, 2008

Student Name:

Marlon Bright

 

Supervisor’s Name and Title at FIU/FAU:

Masoud Sadjadi, Ph.D. – Assistant Professor, School of Computing and Information Sciences, FIU

 

Name of the PIRE International Partner’s Institution:

Barcelona Supercomputing Center (BSC)

 

Supervisor’s Name and Title at the PIRE International Partner’s Institution:

Rosa M. Badia, Ph.D. – Grid Computing Group Manager

 

Project Title:

The Grid Enablement and the Performance Analysis of the Weather Research and Forecasting (WRF) code

 

Problem Statement:

Due to the enormous impact of Hurricanes on all aspects of society, there is a crucial need for the timely availability of precise forecasts and information in order to successfully mitigate the resulting effects.  Forecasts are currently modeled by the Weather Research and Forecasting (WRF) code, which has not been enabled to scale out of a compute cluster environment.  To add to the necessity, the demanding resource requirements of WRF entail a large number of compute nodes, which are often not found in a single organization.  Therefore, it is necessary to enable the WRF code for a grid environment in order to utilize resources in cooperating organizations.  The resulting process is highly complex and prone to inefficiency.  Subsequently, it is, also, essential to develop and implement performance analysis tools that allow application developers to identify and troubleshoot inefficiencies that cause poor performance.

 

Motivation and Impact:

The motivation of this work stems from the need to successfully mitigate the effects of weather catastrophes such as hurricanes.  Currently, forecasts are performed on large scales.  However, there is a need for small scale data in order to model at a higher resolution and yield more accurate and precise local level forecasts.  The impact of this project would lead to more timely and accurate forecasts that can be used by local communities in preparation for disaster in order to be fully prepared for the looming situation.  In addition, this project would advance the technology of Grid Computing in that a highly complex and resource demanding application would have successfully been grid-enabled.  This breakthrough would signify that the required technologies have developed to the point that the Grid can be understood and implemented by other crucial and resource demanding scientific applications.

 

Current Status:

The Grid Enablement of the WRF code is an ongoing research project within the Florida WRF ensemble and the member institutions of the Latin American Grid.  To this point, efforts have been made towards the transparent grid enablement of scientific applications using tools such as TRAP/J and GRID Superscalar.  In this field, applications have been developed to operate across a grid, but still a need exists for a “separation of concerns” between the application source code and the grid enablement code.  This would allow for an ease of applications designed to be executed on a local machine or in a single cluster of processors.  Other work is ongoing here at FIU concerning meta-scheduling and job-flow management in the grid environment.  Also, here at FIU and at BSC  research is being conducted towards the development of application performance analysis and prediction tools to be used for the development and optimization of grid applications.  Two such applications that have been developed here at FIU are amon and aprof, that have been shown to accurately predict the execution time of an application in the grid environment.  Two more tools, Dimemas and Paraver, have been developed at the Barcelona Supercomputing Center (BSC), a PIRE partner institution.  Dimemas is a performance prediction simulator for message passing appications, and Paraver is a visualization and analysis tool for message passing applications.

 

Research Roadmap:

I began researching with my advisor in the Fall of 2007.  At the beginning of this semester, I joined the WRF Grid-Enablement team.  Our current research focuses on developing an effective modeling and profiling method for parallel applications using the tools described above, amon, aprof, Paraver, and Dimemas.  In addition, I am enrolled in a Global Cyber-Bridges course here at FIU entailing the Grid Enablement of Scientific Applications this semester.  Below is a tentative timeline for this effort:

  • Present – April: Research through studying related works and conducting experiments the use of profiling and modeling tools to optimize the grid performance of parallel applications
  • April: Submit paper for publication for findings to Grid ’08 Conference
  • May – June: Continue current research at BSC of the grid-enablement of WRF and its optimization with potential access to Marenostrum Super Computer; continue Grid Enablement course
  • End of June: Submit 2nd paper for publishing on the enabling and optimization of WRF the code as a result of research at BSC  

 

Relation to PIRE Core Research Projects: Briefly explain how the project fits into the PIRE theme of Cyberinfrastructure enablement. Which box in the table (our “9 box” figure at http://pire.fiu.edu/research.php) does your research fit into and why? If this project is not directly related to any of the boxes in the table, then provide a rationale on how this project is related to Cyberinfrastructure enablement.

My research project proposal directly falls in two of the boxes displayed in the figure in two different levels.  In the “CI Applications” the Grid Enablement of the WRF code directly ties in with Hurricane Mitigation Applications.  In the “CI Integration” layer the use and development of tools such as amon, aprof, Paraver, and Dimemas fall into the Visualization Software Tools. 

Posted by Marlon Bright

You must be logged in to post a comment.