Grids Unleash the Power of Many
To paraphrase a wise science officer, the computing needs of the many outweigh the needs of the few. Scientists say new computing grids that put supercomputers in the hands of the masses could exponentially increase scientific discovery and spark new industries.
Computer scientists in three states – West Virginia, North Carolina, and Colorado – are each combining their technology resources into separate computer grids that will give researchers, universities, private companies and citizens access to powerful supercomputers.
The project designers say these information aqueducts will encourage business development, accelerate scientific research, and improve the efficiency of government.
“Grid computing will provide 1,000 times more business opportunities than what we see over the Internet today,” says Wolfgang Gentzsch, managing director of grid computing and networking services at MCNC in Research Triangle Park, NC.
MCNC is spearheading North Carolina’s statewide grid development that currently includes seven universities including North Carolina State, Duke, and the University of North Carolina.
The North Carolina project – which has a goal to link 180 institutions – is encouraging business development through its Start Up Grid Initiative, which allows fledgling companies to plug into the grid for up to nine months free of charge and afterwards at discounted rates, Gentzsch says.
Because raising capital and acquiring technology takes up most of a new company’s time, “Startups usually only get to spend 10 percent of their time executing their idea,” says Gentzch, who has launched seven companies.
According to a 2003 report by Robert Cohen, a Fellow at the Economic Strategy Institute, North Carolina’s grid could create 24,000 jobs and boost the state’s output by $10.1 billion by 2010 if effectively implemented.
Before statewide grids can become a realit, the software used to share and manage resources needs to be improved to include more standard communication protocols. Gentzsch says the expected release of version 4.0 of the open source Globus Toolkit, which he estimates is used by 90 percent of grid projects, will greatly simplify connecting computers to the grid.
Securing a location’s computing resources so that only specified resources are made available for sharing is a significant challenge, Gentzsch says. To protect data files, institutions must “encrypt everything,” and configure the grid network so that “the CPU cycles are separated from the disk resources.”
Gentzsch estimates that advanced computing resource utilization is just 25 percent, and grid computing could increase the efficiency to 75 percent.
The West Virginia Cluster Computing Grid seeks to level the playing the field in academic research, according to Eric Lamar, a principal investigator at the Institute for Scientific Research, which is helping to build the grid.
Lamar says the grid will enable academic researchers who work at smaller rural institutions to gain access to the same quality of resources as West Virginia University.
The grid will increase collaboration between university researchers who share common interests and enable difficult problems to be addressed in concert.
“The more interesting problems of science can be tackled when you have access to more sophisticated computing,” Lamar says.
Lamar says the grid will enable smaller colleges to delve into computer modeling in areas important to West Virginia, such as oil and gas exploration, and medical diagnostics. Grid researchers are currently joining the resources of West Virginia University, Fairmont State University and Wheeling Jesuit University, and three more schools will be added next year, Lamar says.
Bob Marcus, director of Colorado’s CoGrid project, which is just getting started, says that figuring out the business model is a difficult challenge. While everyone agrees that having access to more resources is desirable “in the end, someone has to pay for it,” he says.
The companies and research facilities that have the most money also have sufficient computing resources, while smaller institutions as well as primary and secondary schools don’t have the money to pay for the computer time. State funding can get the ball rolling, but eventually commercial applications must be developed, according to Marcus.
Just as the Internet served as a platform for developing new classes of applications (Web services, peer to peer networks), beyond what its inventors could have imagined, Marcus sees grid computing as a testbed for experimenting with innovations to come.
Grids would likely grow from state projects to regional initiatives based on common needs. For example, the mountain states may want to share resources to track livestock movement, or western states could join together on simulations for wildfire response.
“We may have to kiss a few frogs before we find a prince,” Marcus says.
Statewide grids must be designed to guarantee that any organization that contributes a resource can gain control of that resource when needed, according to Marcus.
“If you paid for it, you own it,” Marcus says, otherwise organizations won’t sign on.
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today