The National Science Foundation has awarded an additional $35 million to three research institutions to expand the TeraGrid project, a multi-year effort to build and deploy the world's largest, fastest and most comprehensive distributed infrastructure for open scientific research.
The National Science Board approved the awards that will be made by the NSF, on top of $53 million awarded to the TeraGrid project a year ago. Final amounts will be determined through negotiation between the NSF and the awardees. The NSB is a 24-member policy body that guides the NSF in its role as an independent federal agency supporting scientific and engineering research.
The Extensible Terascale Facility award expands the TeraGrid to five sites: the San Diego Supercomputer Center (SDSC) at the University of California, San Diego; the National Center for Supercomputing Applications (NCSA) at the University of Illinois, Urbana-Champaign; Argonne National Laboratory near Chicago; the Center for Advanced Computing Research (CACR) at the California Institute of Technology, Pasadena; and the Pittsburgh Supercomputing Center (PSC) at Carnegie Mellon University and the University of Pittsburgh.
This extended TeraGrid environment will provide the national research community with more than 20 teraflops of computing power distributed among the five sites and nearly one petabyte (one million gigabytes) of storage capacity.
The award also ensures that the TeraGrid will be extensible and ready for expansion in the future. Additional sites will be able to connect to the TeraGrid, and the national research community will be able to take advantage of its high-performance resources.
SDSC, NCSA, Argonne, and CACR were already part of the three-year, $53 million TeraGrid project, announced a year ago by the NSF. PSC had previously received a $45 million NSF award to build a terascale computing system, called TCS-1. The ETF award integrates these two efforts and their computing environments to create an extended terascale-level Grid of data, computation, and visualization resources that will make possible new scientific discoveries. The five sites will be linked by the world's fastest dedicated optical research network, built in partnership with Qwest Communications and designed to accommodate additional connections.
"With the ETF award, the TeraGrid has evolved to encompass even more powerful terascale computing and data management facilities, forming the basis of tomorrow's national information infrastructure," said Fran Berman, director of SDSC and chair of the TeraGrid executive committee. "The TeraGrid has begun to grow in both size and heterogeneity, and other sites will be able to connect to it in the future. The sheer number, scale, and diversity of resources promises enormous scientific potential in the first decade of the digital millennium."
"This expansion of the TeraGrid will provide computing power to scientists that is orders of magnitude beyond anything we've ever seen before," said Dan Reed, director of NCSA and chief architect for the TeraGrid project. "In addition, it will provide the best high-resolution visualization environments, more storage capacity than ever before possible, and access to Grid computing toolkits and Grid-enabled applications. The impact on scientific discovery will be significant. At this point, we can't even begin to imagine the discoveries that the TeraGrid will make possible."
Argonne and NCSA are already connected by network links of 20 gigabits per second (Gb/s). By the end of this fall, Qwest Communications will connect the two Illinois TeraGrid sites with the two California sites through a central backbone network running at 40 Gb/s between StarLight, the optical interconnect based in Chicago, and the major Internet hub in Los Angeles. PSC will connect to the central backbone network at 30 Gb/s through StarLight.
"By combining TCS-1 and the TeraGrid, this award will create the first wide-area Grid encompassing terascale systems of differing architectures," PSC scientific directors Mike Levine and Ralph Roskies said in a joint statement. "This heterogeneity, which results from linking Pittsburgh and the TeraGrid, will become the basis for a realistic generalizable Grid infrastructure."
Each of the five sites will play a specific role in the TeraGrid project.
SDSC will serve as the lead data and data services site, providing 500 terabytes of storage area network (SAN) disk storage augmented by two 32-processor IBM Power4 DB2 servers joining the existing 72-processor Sun Fire 15K server. The site also will deploy a four-teraflop Linux cluster based on Intel's 64-bit Itanium processor and a 1.1-teraflop Power4 computing system.
NCSA will be a lead computing site providing more than 10 teraflops of computing power in Linux clusters equipped with Intel's second and third generation 64-bit Itanium chips. NCSA will also offer 230 terabytes of SAN-based storage.
PSC will be a lead computing site providing a six-teraflop TCS-1 system and a 0.3-teraflop HP Marvel system, creating a powerful shared memory computing capability with 512 gigabytes of memory as well as adding a 150-terabyte disk cache.
Argonne will serve as the TeraGrid's remote rendering and visualization site. Its 180-node visualization cluster will be the largest such cluster for open scientific research.
CACR will deploy a data collection and analysis environment centered around a Sun Microsystems data server. The system will be a prototype of a mid-size, application-driven TeraGrid site.
The new equipment will be installed and deployed over the next year and a half. Researchers from across the country will be able to tap into the TeraGrid's resources through high-performance research networks, such as Internet2's Abilene network. The NSF funding will go to SDSC, NCSA, and PSC. Argonne and CACR will receive their funding through the three awardees.
TACC Awarded $2 Million by Department of Energy for Grid Web Services
The Texas Advanced Computing Center (TACC) has been awarded a three year grant totaling $2.1 million by the Department of Energy Mathematics, Information, and Computational Sciences (MICS) program to support the DoE Scientific Discovery through Advanced Computing (SciDAC) Collaboratory projects through the development of new Web portal technologies.
"The overall goal of this project is the development and deployment of interoperable portal and Web services that can be used by a large number of independent users across the entire DoE Science Grid," said Mary Thomas, TACC Grid Computing Group Manager and Principal Investigator for the project.
Other members of the team, who will receive portions of the grant based on their contributions, include co-principal investigator Geoffrey Fox from Indiana University, and co-investigators Reagan Moore of the University of California at San Diego, Dennis Gannon of IU, and Dave Schissel of General Atomics.
The recently funded SciDAC program promotes the advancement of scientific and computational problem solving capabilities for the DoE Office of Science. Within this program, three areas are defined: Scientific Challenge Codes, Computing Systems and Mathematical Software, and Collaboratory Software and Infrastructure. Collaboratory projects will build a uniform support environment in which DoE SC users will be able to run jobs and experiments and to manipulate data on the Grid.
Grid portals provide the scientific community with familiar, simplified interfaces to the Grid and Grid services, with the SciDAC program bringing the Grid to the DoE community, there will be a need to deploy Grid portals onto the SciDAC Grids and Collaboratory. Grid portals are now being used on production grids for large organizations such as the National Science Foundation Partnership for Advanced Computational Infrastructure, National Aeronautics and Space Administration Information Power Grid, and National Institutes of Health Biomedical Informatics Research Network. The software used to build these portals has been converted to generalized application portal development toolkits that simplify the application developers' task of accessing the complex Grid technologies used for Grid services.
The emergence of web service technologies has provided the Grid and portal communities with an opportunity to develop interoperable protocols and standards that can be used for grid services and portals. This will simplify the use of Grids and Grid technologies and will encourage the use and deployment of scientific applications and experiments on the Grid.
The efforts of this team will concentrate on research, development, and deployment activities within four primary tasks: the development of portal systems, management of data collections, migration of technologies from the DoE sponsored Common Component Architecture project, and the development of Web services in support of the above activities. Initial deployment of these technologies will be on the National Fusion Collaboratory, with plans to explore deployment onto the DoE Science Grid, the Collaboratory for Multi-Scale Chemical Science, and other collaboratories in the later phases of the project.