SCIENCE & ENGINEERING NEWS
West Lafayette, IND. — A system at Purdue University could help create a worldwide “computational grid” in which individual users no longer have to purchase software but are able to run programs remotely over the Internet.
The system is called the Purdue University Network Computing Hubs, or PUNCH, a network computer that provides access to programs from 16 universities, four research centers and six companies.
“We had a million hits the first half of this year, and we don’t advertise,” said Mark Lundstrom, a professor of electrical and computer engineering.
PUNCH is now primarily used by engineers who require highly specialized software for research and teaching. The software is not commercially available and is difficult to use and install. PUNCH not only makes the software accessible to the entire research community, it automatically enables computer users to run the software via their own computers through the World Wide Web.
“The software does not actually run on the users’ computers, it runs on a server somewhere,” said Nirav Kapadia, a senior research scientist responsible for developing the underlying software that makes PUNCH possible. “It enables whatever you are running on your server to interface with a distant computer.”
The system has saved money for engineering students by eliminating the need to buy expensive software for certain courses.
“PUNCH is essentially a system that enables users throughout the country and the world to access our computer-based tools, and more importantly, to actually run them through their own computers,” said Jose Fortes, a professor of electrical and computer engineering who worked with Kapadia to develop the system.
Although PUNCH is now dominated by engineering applications, in principle it could be used for a much broader range of software, including programs used for business and industry. Companies with offices in different states or countries would benefit from using such a hub to share expensive software.
“In the long run, I think a grid should serve all computing needs,” Kapadia said. “For now, the research community provides a nice user base to test the ideas behind the grid. I see PUNCH as a prototype of a grid; its purpose is to allow us to learn what needs to go into building a large-scale infrastructure of this type. Personally, I would love to see PUNCH evolve into a commercial computational grid.”
However, developing a commercial grid would pose challenges. “The software vendors are going to have to sort out how to license and charge for this service because the implication is that people would not have to buy software,” Lundstrom said.
The grid concept has recently been attracting more interest from corporate America. “This has suddenly, within the last year or so, become a very hot field,” Lundstrom said. “Companies called application service providers are beginning to jump into this. The advantage that we have is that we’ve been doing it for five years now. We’ve learned what it takes to make it work.”
An important feature of the system is that it can automatically find resources, anywhere in the nation, that users need to do their work.
“If you are running a high-level simulation and it requires a super computer, it can look around for a super computer that isn’t being used at the time, or one that is being least used, and send the job there,” Lundstrom said. “If you run a smaller job, it can recognize that it can go to a smaller machine somewhere else. It can manage all of this, like the electric power grid. Your electricity comes from someplace where there is excess capacity, and it is routed across the country. In the same way, there are a lot of excess computing cycles that just aren’t used that could be used somewhere else.”
PUNCH contains various “hubs” for different engineering interests, and each hub connects users to the programs they need for their work. For example, Purdue engineers have most recently created a nanotechnology simulation hub, or nanoHub, which provides programs for designing extremely small transistors and other components measured in nanometers, or billionths of a meter.
“We are trying to convert computing into a service infrastructure,” said Kapadia, who began developing the system to help Lundstrom solve a research dilemma.
“I had some high-level simulation tool that I wanted to share with my colleague in another state,” Lundstrom said. “He wanted to hire a post-doctoral fellow to install this and to learn how to run it.”
Within a half-hour Kapadia had the complicated simulation program on the Web. “I called this fellow back in New York and said, ‘You don’t need to acquire the software, just log onto this address and run it,’ and he was doing that in the afternoon,” Lundstrom said. “Within a couple of months we had written a paper together. It probably would have taken him a year to have gotten the software, converted it to his machine, and learned how to run it himself.”
Kapadia and other members of the PUNCH team presented details about the system in November during an international computer conference, SC2000, sponsored by the Computer Society of the Institute of Electrical and Electronics Engineers.
Lundstrom has developed a simulation tool for designing transistors only a few atoms wide, and in December he made the tool available to the research community via the nanoHub. The simulation tool has yielded information demonstrating the benefits of a new type of transistor. Purdue researchers say they know of only two other teams in the world that have created such a simulation tool. But those programs are not accessible to the research community at large.
“We are going to make this available to anyone,” Lundstrom said.