Shutdown Of Supercomputer Firm Imperils Princeton Installation

The John von Neumann Supercomputing Center at Princeton University may be paying a devastating price for its alliance with a floundering computer giant. Last month the four-year-old center was rocked by the announcement thai its sole supercomputer supplier, ETA Systems Inc., had been shut down abruptly by its parent company, Control Data Corp. (CDC). That decision, made after the six year-old supercomputer subsidiary lost more than $100 million last year, means the center will get no further hardware upgrades and only minimal software support from CDC in the future. The news couldn’t have come at a worse time for the von Neumann center, which is run by a consortium of 13 research universities, has a user base currently numbering more than 1,300 university and industrial scientists, and serves as the hub of a recently completed high-performance regional network. This month the National Science Foundation, which provides a total of $54 million each year to Princeton and four other centers around the country, is expected to decide whether to renew its commitments for another five years. The other four centers—housed at Carnegie-Mellon University, Cornell University, the University of Illinois, and the University of California, San Diego—fully expect to receive a vote of confidence from the National Science Board. (With the exception of the Cornell center, which uses IBM supercomputers, all the others use Crays.) But Princeton’s problems are seen as a crippling blow to its chances for a new round of support. “They’re in real desperate straits,” says John Connolly. of the University of Kentucky, who oversaw the von Neumann center’s launch in 1985. “They’ve got to come up with something quick to convince NSF that they should be renewed.” William Wulf, NSF associate director for computer and information science and engineering, says he has assured von Neumann center officials that the agency "Won't act precipitously.” But, he noted, the center “needs to think creatively about what to do” about its newly orphaned computers and its rapidly dwindling software support. At the very least, Wulf adds. Princeton needs to demonstrate that it has a realistic plan to upgrade its current machines. NSF officials have asked the science board for a three-month extension so the center can rethink its future and submit a new proposal. But officials are short on encouragement for the center. “Its options are limited at this point,” admits Wulf. From the start, the center’s decision to go with CDC’s financially strapped supercomputer subsidiary was a controversial one. The von Neumann center is built around two ETA 10 supercomputers and two CYBER 205 high-end mainframes, also made by CDC. But progress toward maximizing the center’s computing potential has been blocked by the hardware supplier’s financial woes, chronic delays in delivery of new basic operating system software, and poor industry acceptance of the machines. In hindsight, says Wulf, it’s clear that Princeton’s 1984 decision to choose ETA computers over those made by its larger rivals, Cray Research Corp. and IBM Corp., was a mistake. “If we had known that CDC support was going to be so poor we probably would have taken a different tack,” he says, noting that congressional pressure to support U.S. supercomputer manufacturers played a part in the original decision. In the past few years, the von Neumann center appeared finally to be getting on its feet, with a strong show of support from other universities. But the unexpected CDC announcement changed everything. “The timing couldn’t have been worse,” notes Connolly. “They had their proposal [for funding renewal] all wound up in a neat package and bow, ready for the science board.” The center currently receives $12.5 million annually from NSF. Doyle Knight, president of the von Neumann center, calls CDC’s decision to close down ETA “disappointing.” But, he says, he needs to talk to CDC about their contracts before deciding what to do. CDC has said it intends to “continue to suport” current ETA customers, he notes, but a company spokesman explains that the terms must be negotiated on a case-by-case basis.. If NSF support does indeed diminish or disappear, one possible solution for the Princeton center lies in the prospect of increased funding from the state of New Jersey, whose Commission on Science and Technology this year is providing the center with $2.45 million. Doyle says that both the commission and Sen. Frank Lautenberg (D-NJ.) “have been informed” of the center’s plight, but that no action is yet planned. Lautenberg sits on the Senate appropriations subcommittee that determines NSF’s funding. The fierce competition for NSF’s limited funds is also likely to be a factor in any decision. As Connolly notes, “the other four centers are like sharks smelling blood. They’re looking for [money for their own] upgrades, and they’ll use this as their opportunity.” Larry Smarr director of the competing University of Illinois National Center for Supercomputer Ap- plications, admits that his renewal proposal calls for a major upgrade (the addition of a massively parallel Connection Machine supercomputer), but he notes that the proposal was submitted before the ETA decision. Even if the von Neumann center is cancelled, he points out, it is not clear that the money saved would go to the other centers. With fewer centers, he says, “Congress could easily cut the whole program instead.” The existing equipment at the von Neumann center will remain competitive for the next three or four years, says Wulf. But he worries that programmers will stop writing software for the ETA machines immediately, if they perceive that there is no future in that endeavor. As with all computers, the ETA machines are only as good as their software, and with no further improvements on the horizon, Connolly adds, “you’ve got to start looking for an alternative immediately.” The ETA pullout will have little immediate impact on the financial condition of the von Neumann center. Because installment payments on the supercomputers—which cost between $30 million and $40 millionare based on their passing certain performance milestones, the center had paid ETA only about $700,000 when CDC folded the company. The real cost to the center explains Wulf, involves the acquisition of new, state-of-the-art machines that would have to be brought in to replace the ETA computers. On the most optimistic side, the ETA debacle can be seen as a chance for the center to rethink its whole approach to supercomputing—and this could prove to be the silver lining in an otherwise bleak cloud, says Wulf. Even without NSF support, the center could begin anew with a radically different-type of computer, providing it can find enough outside funding. “It’s an opportunity to stand back and ask what’s best for the this could prove to be the silver lining in an otherwise bleak cloud, says Wulf. Even without NSF support, the center could begin a new with a radically different type of computer, providing it can find enough outside funding. “It’s an opportunity to stand back and ask what’s best for the country’s scientists,” Wulf says. One possibility, he notes, is a shift toward massively parallel supercomputers, a direction that would set Princeton apart from the other NSF-funded centers and move it into one of the fastest growing fields in computer science.

Shutdown Of Supercomputer Firm Imperils Princeton Installation

Interested in reading more?

Become a Member of