A view of one of the SC|05 exhibit halls from the top of the NCSA booth. |
Grid computing has come a long way since the first "grid," called I-WAY for Information Wide Area Year, debuted at the 1995 Supercomputing conference. I-WAY linked a dozen high-performance computing centers and advanced visualization environments at speeds of up to 155 megabits per second, and was the only grid around. This week at the Supercomputing 2005 (SC|05) conference, SCinet is providing almost one-half of a terabit per second of network connectivity to the show floor, research grids from the Americas, Europe and Asia are on display everywhere you turn, and the Grid Workshop has grown so large that it will soon become a full-fledged conference separate from Supercomputing.
SC|05, the international conference on high performance computing, networking and storage, has taken over downtown Seattle. SC|05 posts some impressive statistics: Over 9,250 registered attendees, requiring 1,300 gallons of coffee and 20,000 beverages; thousands of exhibitors from research and industry; over 30% of the 165 research exhibits showcase grid computing; 100,000s of hours of research have gone into the technical papers being presented; 418 people attending the Grid Workshop; and over 140 high school and undergraduate teachers are attending this year's education program.
Grid computing at SC|05 was kicked off Sunday afternoon by the Grid 2005 Workshop keynote address. Dennis Gannon from Indiana University spoke about lessons he's learned while building a series of grids and reviewed the many challenges still to come in grid research. After discussing early projects such as the I-WAY and NASA's Information Power Grid, Gannon described the new Linked Environments for Atmospheric Discovery project.
The Washington State Convention and Trade Center in Seattle, the venue for SC|05. |
"LEAD is a grid designed to change the paradigm for mesoscale weather prediction," said Gannon. "Our ability to predict hurricanes and tornados is horrible. The old way is entirely serial, static and pre-scheduled, and doesn't respond to weather." LEAD will be an adaptive cyberinfrastructure, in which data from Doppler radar will feed in to hurricane and tornado simulations, which will in turn feed directly back to the radar to collect the data essential for predictions. As the weather changes, so will the services and the scenarios executed by the cyberinfrastructure.
As the infrastructure continues to be constructed for this two-year-old project, LEAD researchers will face many of the challenges Gannon described for grid computing in general, including building a truly lightweight grid infrastructure, interoperability between the many infrastructures that now exist, and easily identifying the data, applications and services that will work for different researchers. For the latter, Gannon suggested a "Google" for grids—a search engine that would allow researchers to easily find the data, applications and services on the grid that are of interest to them.
The workshop continued with a series of presentations on topics from data replication to debugging, and concluded with a panel discussion on the future of grids that featured Fran Berman from SDSC, Fabrizio Gagliardi from Microsoft, Carl Kesselmann from USC/ISI and Mark Linesch of GGF. Gagliardi suggested that in five years, grids will essentially disappear, becoming just another part of the infrastructure. Other topics of discussion included funding models for grid research, grids in the commercial and gaming sectors, adoption of standards and the role of virtual organizations.
Learn more at the SC|05 Web site.
—Katie Yurkewicz
e-mail this article
|
