TeraGrid: A Prototype Cyber-Infrastructure...

Technical and Organizational Lessons

The TeraGrid project, funded by the National Science Foundation, is nearing completion of the deployment of a high-performance Grid system with resources located at Argonne National Laboratory, Caltech, the National Center for Supercomputing Applications, the Pittsburgh Supercomputing Center, and the San Diego Supercomputer Center. With a total investment to date of approximately $100M, the TeraGrid includes over 20 Teraflops of computational resources, over 1 Petabyte of online storage, and a 40 Gigabit/second interconnection network. A project of this scale involves critical issues not only in technology integration but also in organizational processes, project management strategies, and close coupling with "customers" to ensure that the overall system delivers value to the science and engineering community commensurate with the investment. As importantly, an examination of the results of NSF's TeraGrid investment strategy and relative technical benefits of building a distributed capability versus stand-alone resources, or a single stand-alone resource yields important insight to guide investments today and into the future with respect to high performance technology investment on a national scale.