With this project, Louisiana State University (LSU) joins the integrated, national cyberinfrastructure known as the TeraGrid, contributing high-performance computing (HPC) services and grid technology expertise to support the nation's science and engineering researchers. This project extends the computational capacity of the TeraGrid by providing access to 50% of a Dell Linux cluster that employs Intel's multi-core processor technology. In addition, the project contributes access to several data management and grid services. These services include the PetaShare distributed data archival, analysis, and visualization system, the Highly-Available Robust Co-allocator (HARC) co- scheduler for compute, network, and data services, and the Cactus Computational Toolkit. The project includes the provision of user support, with an emphasis on support for researchers wanting to develop grid applications.
The project includes a statewide outreach program anchored by Southern University, a large HBCU system. It is anticipated that the project will leverage LSU's connections with the Louisiana Optical Network Initiative (LONI) and the Southeastern University Research Association (SURA) to increase the usage of TeraGrid by under-represented groups. Project staff will work with researchers in hurricane and storm surge modeling to develop a TeraGrid Science Gateway for event-driven, ensemble modeling of storm surges when hurricanes approach US coastal regions, a project with the potential to contribute to efforts to improve national hurricane preparedness.
The net effects of this award will be: (i) to provide the science and engineering research and education community with access to a high-performance computing resource with a familiar and popular architecture; (ii) to enhance the expertise in and user support for grid services within the TeraGrid; (iii) to increase the geographical diversity of the TeraGrid partnership; and (iv) to enhance the TeraGrid's connections with universities that include significant numbers of students and faculty from under-represented groups.
The funded HPC services will permit investigators across a broad range of disciplines to conduct advanced research on topics in many areas of science and engineering. It is anticipated that this award, together with other recent awards in NSF's high-performance computing program, will collectively strengthen the provisioning, to the science and engineering community, of high-end cyberinfrastructure services - such as high-performance computing, community data collections and data analysis services, remote visualization, and science gateway services - through the nationally distributed cyberinfrastructure system that is the TeraGrid.
Description As a major component of the HPCOPS (High-Performance Computing Operations) grant, the Louisiana State University (LSU) and the Louisiana Optical Network Initiative (LONI) contributed computing services to the National Science Foundation (NSF) TeraGrid (the Extensible Terascale Facility) user community as one of eleven resource providers. The provided system included 334 computational servers in a cluster totaling 668 machines. It made use of the LONIâ€™s 40 Gbps optical network to provide high-speed data exchange with other TeraGrid systems. In addition, specialists in computational techniques provided user support services via a help desk run by LSU for all the high performance computing systems it was responsible for operating. Besides answering user questions, this group of specialists actively participated in user training via frequent tutorials, specialty courses, and workshops. Many of the training sessions were presented via Access Grid, a remote educational system which provided real-time video and audio sessions, to users around the state and across the country. The HPCOPS project initiated a national presence at LSU/LONI in supercomputing and cyberinfrastructure, which has allowed the LSU and Louisiana user communities to expand into the national computational science community through direct support and development activities. Major Products DARE-SAGA Modern computer systems are not exactly easy for scientists to program. The amount of computer science training and hardware knowledge needed to do it correctly and efficiently consumes time and effort that might be more productively spent on the actual science study. LSU has coupled a programming framework, called Cactus, with a web-based user interface, called the DARE-SAGA (Dynamic Application Runtime Environment - Simple API for Grid Applications) web portal, to make it much easier to run complex problems. Cactus allows programmers to focus on writing kernels which describe block-shaped regions of a problem. The framework can then call such blocks repeatedly to create parallelism, to use small blocks to study a sub-region in detail, use multiple blocks to cover an oddly-shaped region, implement different solutions methods, or run on specialty hardware such as graphics processing units (GPUs). The web interface makes things even simpler, as it allows the scientist to design a new program using the building blocks, and will even send it off with data files to be executed without need for creating job control information. BigJob While some scientific studies concentrate a tremendous amount of computer time on a single problem, others have 10's of thousands of small problems that may collectively require just as much time. Each of these small problems has their own input, output, and control files, turning the analysis into a huge book-keeping problem just in organizing and controlling the work. BigJob was developed to solve the book-keeping problem and allow the scientist concentrate on results. BigJob works with a set of locally developed tools, called SAGA, which controls job submission to a variety of job processing systems. Working together, SAGA and BigJob frees the scientist from having to deal with the various methods of running jobs on different systems, and provides a data management service to submit, execute, and collect results from thousands of jobs. Other The unique cyberinfrastructure capabilities developed under this grant allowed LSU and LONI to provide crucial national disaster help in the hours following the 2010 Deepwater Horizon oil spill in the Gulf of Mexico. The National Oceanic and Atmospheric Administration (NOAA) was provided emergency access to the LONI high-speed, high-bandwidth networking connections allowing critical data to be quickly shared between New Orleans and NOAA offices. This gesture increased the ability of first-responders to target aid where needed most. LSU and LONI researchers also utilized the advanced computational capabilities to help inform the response teamâ€™s understanding of the oil spillâ€™s impact in various areas. During hurricane seasons, LSU and LONI participated with the broader coastal modeling community to assist the state and the US Army Corps of Engineers with storm surge predictions whenever a storm was predicted to approach the US coastline. Such predictions helped determine the risks faced by different areas. The information produced facilitated disaster relief planning and response.