NSF funds IU’s ‘Global Grid’

By Christine Fitzpatrick

This “Global Grid” of computing networks will harness the computational power needed to conduct major scientific experiments in physics, astronomy, biology and engineering in the 21st century.

The iVDGL will consist of a seamless network of thousands of computers at 40 locations in the United States, Europe and Asia. Together, these computers will work as a powerful grid capable of handling petabytes of data. One petabyte is roughly the amount of data contained in 100,000 personal computer hard drives.

Software developed by the GriPhyN (Grid Physics Network) project—a major NSF computer science-physics consortium funded in 2000 in which IU is also involved—will allow grid users to access and process globally distributed data in a transparent fashion. The iVDGL will be the platform on which these grid technologies can be evaluated and developed.

IU will make significant contributions to the iVDGL by providing two major components of the grid computing project: a prototype Tier-2 Data Center for the ATLAS high energy physics experiment and the International Grid Operations Center, or iGOC. The project makes use of investments made by IU in several key areas of information technology, including advanced networking (Internet2 Abilene, TransPAC and Global Network Operations Center), high performance computing and massive data storage.

The prototype Tier-2 Center will be a data analysis facility for physicists using the ATLAS detector at the CERN Large Hadron Collider, located near Geneva, Switzerland. Approximately 2,000 ATLAS physicists will seek to understand the origins of mass and will search for new forms of matter and interactions at the high energy physics frontier starting in 2006.

The U.S. component of the ATLAS team consists of researchers from 29 universities and three national laboratories. The Tier-2 Center, located on the Indiana University-Purdue University Indianapolis campus, will house “farms” of Linux-based computers, which will process data loaded onto large arrays of hard disks and the university’s high performance tape storage systems.

These computer farms will be connected to others at centers around the world via the Internet2 Abilene network, and will be used by physicists to simulate particle collisions in the ATLAS detector in a series of “data challenges” beginning next year. The ATLAS experiment, which will operate for a decade or longer, will begin producing data at a rate of roughly 10 petabytes per year.

“We are helping to develop the international computing platform required to extract physics measurements from the ATLAS experimental data,” said Robert Gardner, assistant professor of physics at IUB, who leads the distributed IT infrastructure effort for the U.S. ATLAS Collaboration and the IU GriPhyN and iVDGL teams. “The enormous amount of data produced by ATLAS will be distributed in a hierarchical way to many data centers in Europe, the Americas and Asia. Grid software will make it easy for physicists to use specific data collections no matter where the data is physically stored. In addition, results from their calculations will be accessible to collaborating physicists located throughout the world.

“The roadblocks we will remove while building the software infrastructure for an internationally integrated data grid facility will benefit other large-scale, data-intensive sciences at IU such as astronomy, astrophysics, bioinformatics research and earthquake seismic imaging,” Gardner said. “Our applications share many requirements, the most important being the ability to securely and efficiently locate, access and process large data volumes distributed over the network by authorized researchers located anywhere.”

The effective operation of this globally distributed grid laboratory requires certain coordinated support services and management. This support will be provided through the development of an International Grid Operations Center (iGOC) to be co-located with IU’s Global Network Operations Center (NOC) on the IUPUI campus.

“The iGOC will support the iVDGL as the Global NOC supports international research networks, providing a unified and coordinated point of contact for iVDGL status, configuration and management, as well as overall issues of robustness,” said Brian D. Voss, IU associate vice president for telecommunications.

A separate but related NSF grant was awarded to IU to develop software to acquire “telemetry” signals from the grid for operational use by the iGOC. This three-year program will result in software systems used by the iGOC to monitor and archive grid status and performance data, and make these data available to developers and users of the grid. The NSF iVDGL and grid telemetry awards will provide $2.4 million in funding to IU over the course of the two projects.