IBM and the U.S. Department of Energy announced on Friday that they have begun building a nationwide computer grid that ultimately will be capable of processing more than 10 trillion calculations per second and storing information equivalent to 200 times the number of books in the Library of Congress.
The grid eventually will offer scientists and researchers across the country real-time access to trillions of bytes of data stored at labs nationwide, according to IBM.
Computer grids are designed to allow quick access to applications, data and computing resources housed at distant locations. While similar in some ways to the Internet, grids offer far greater advantages by enabling users to tap the computing power of potentially thousands of systems attached to the grid simultaneously, creating a virtual supercomputer.
Efforts to develop grids are underway in various countries throughout the world. Last summer, Britain announced plans to build a national grid linking nine research centers and the National Science Foundation announced it would link four of its U.S. supercomputing centers into a power grid network.
A major challenge to developing grids centers on creating common protocols to enable computers utilizing different proprietary hardware and software to not only communicate with each other, but interact almost as closely as if they were all using one operating system.
The IBM and DOE grid will be based on clusters of servers joined together over the Internet using protocols developed in conjunction with the Globus open source community, as well as other open source technologies, such as Linux.
While most current grid projects are primarily focus on aiding government and scientific research, IBM representatives said the technology could eventually find its way into commercial use.
“The DOE Science Grid is a template for the kind of system that can enable partnerships between public institutions and private companies aimed at creating new products and technologies for business,” said Val Rahmani, general manager of IBM eServer, pSeries.
Eventually, IBM, Armonk, N.Y., believes there will come a time when companies and private individual may actually purchase computing power in a way not that unlike how customers currently buy electricity, essentially paying for what they use.
“This collaboration is a big step forward in realizing grids promise of delivering computing resources as a utility-like service,” Rahmani said.
In taking the first steps to build out the grid, IBM and the National Energy Research Scientific Computing Center (NERSC), a part of the Department of Energy, said Friday that they have successfully integrated the first high-performance computers and storage devices into the grid.
Among the systems connected was a 3,328-processor IBM supercomputer used by the NERSC. That system is listed as the third most powerful computer in the world, according to the TOP500 List of Supercomputers.
In addition to the large supercomputer system, grid software will be integrated into NERSCs High Performance Storage System archival data storage system, which has a capacity of 1.3 petabytes and is managed using IBM servers.
The primary goal of the new grid, according to the DOE, will be to enable scientists at national laboratories and universities around the country to perform ever-greater calculations, manage and analyze ever-larger datasets, and perform more complex computer modeling necessary for DOE to accomplish its scientific missions.
The NERSC is located at the DOEs Lawrence Berkeley National Laboratory in Berkeley, Calif.