Duke University Alumni Magazine

BUGGED BY THE MILLENNIUM Y2K BY MONTE BASGALL Illustration by Walter Stanford Are the computer devices most societies depend upon primed for dysfunction when the clock turns 2000? If so, we may be facing what Byte magazine calls "a crisis without precedent in human history."

he most symbolic New Year's Day in ten centuries may take on a disaster-movie flavor thanks to one of the greatest blunders in computer technology's short history. Just picture a December 31, 1999, where the lights go out at the end of the countdown on New York's Times Square. What if the international TV audience watching the ball drop see their screens suddenly blacken, shutting off the strains of Auld Lang Syne? What if traffic lights, elevators, electric-power grids, telecommunications networks, and air-traffic control systems also suffer disruptions? What if cars suddenly refuse to start and hospital medical monitors stop making measurements? As a follow-up, what if worldwide stock markets suffer gridlock on January 3, 2000 --the next thousand years' first business day?

As one, arguably sensationalized, Internet article warns: "Unless they are fixed, all computer programs, everywhere in the world, will go on strike on January 1, 2000." Unfortunately, the author, Peter de Jager, who bills himself as an "industry speaker on the topics of change, creativity, and management technology," isn't kidding. Experts agree that at least some of this horror script could come true--though fortunately only the far fringe think it all will--due to a global planning debacle of epic proportions variously called the "Year 2000 Problem," the "Millennium Bug," or "Y2K." Simply put, many of the computerized devices most societies depend on have long been primed to dysfunction at the stroke of the Millennium. And it's all because they use calendars that cannot accommodate an extra digit. This time bomb has not only been widely programmed into countless software programs, but it has also been burned into the very structures of microchips embedded deep inside a wide variety of electronic devices. Since computer hardware and software are the engines of the industrialized world, the potential repercussions are mind-boggling.

"If it has electricity going through it, it is suspect," says William Edge, IBM's Year 2000 consultant and solutions manager. "I think we all agree that no problem facing us is more pressing, especially since, unlike other Washington problems, neither the president nor Congress can push the deadline back," added former Duke trustee chairman John Koskinen '61, who is now the White House's Year 2000 czar, during his testimony to a U.S. House of Representatives subcommittee.

Byte, a computer magazine, called Y2K "a crisis without precedent in human history. We know exactly what's causing it and what to do about it." So, with just over one year left, can the problem be fixed in time? Probably not, Byte opined, at least not completely.

Such pessimism was also in evidence at the first of two "Year 2000 and the Millennium Bug" conferences for the Duke community organized by Duke's Office of Information Technology (OIT) to spread the word that the Millennium Bug is real. Serving out a mixture of humor and alarm, David Kirby used terms like the "Lilliputian" and "Titanic" effects to describe potential scenarios as the calendar fatefully flips. The Lilliputian Effect, a reference to Gulliver's Travels, means "many small problems at once" bogging down operations; the Titanic Effect, a reference to 1998's largest-grossing disaster movie, means "everybody in trouble at once," he said.

Growing more serious, Kirby, who is Duke Medical Center's Year 2000 coordinator, reported findings that about 5 percent of the medical center's 20,000 clinical devices would have failed were nothing being done to correct their Y2K-related chip glitches. Some of those problems "would have killed people," he warned the audience. IBM's Edge later added, "Before this problem is resolved, it will probably touch some facet of everyone's life in this room."

At that conference, and at another held three weeks later, anybody who didn't know already learned that the Millennium Bug dates back to the 1960s and Seventies, when computers had far less memory storage than today's. To save vital space, programmers of the time decided to squeeze in only the last two digits of four digit years, entering 1971, for example, as "71." This abbreviated shorthand records the year 2000 as "00." And, having far less insight than people, computing systems will interpret that to mean 1900--all over again. If you turn on your home computer on New Year's Day, 2000, everything may seem normal until you try to read your electronic mail. Because all e-mail is "date stamped," your suddenly confused computer may file it in a strange place if it thinks the year is a century ago, says Tom Noffsinger, a Vanstar Corporation Year 2000 consultant to Duke's Medical Center Information Systems office.

If you're lucky, the message may simply be placed at the very bottom of your e-mail list so that you'll only have to "move the cursor down" to find it, Noffsinger says. If you're unlucky, "it could go off into 'bit land,' out in the middle of nowhere, where no one ever sees it again."

If it's January 3, 2000, and you're working on an office computer that is tied to others, you may face another set of problems. Not only must the computers' calendars all be synchronized, so must the separate software system that runs your computer network. What happens if everything is not in perfect sync? "We have a pretty good idea in specific cases, but we don't in general know what may or may not happen," Noffsinger admits. "I don't think airplanes are going to fall from the sky. But we don't know where your word-processing file that you've got to get done by a deadline is going to end up." In big mainframe computers, the kind that handle everything from billing invoices and salary records to doctors' appointments and student loans, losing track of the date could obviously wreak havoc. At best, the results might be bills or paychecks bearing the wrong information. At worst, the date confusion might provoke a computer to crash. If that happens, it may be especially difficult to find someone to fix the damage quickly, because a lot of computers may be crashing simultaneously on that memorable New Year's Day. Remember the Lilliputian Effect and the Titanic Effect.

While many Y2K planners initially focused their concerns on software, they're now equally nervous about the danger from "embedded" hardware. That means two-digit dating that has actually been "burned into a chip" rather than being written into software code, says Nevin Fouts, the associate dean for information technology at Duke's Fuqua School of Business: "It's not meant to be changed without replacing the whole chip. And the chip might be three levels inside of some device that you can't even take the case off to try to find. We can connect into a software program, and we can then see the code, edit it, and change it. But with embedded systems, if you can't get to it, you're not even sure of where the logic is."

Fouts says that the instructions etched into embedded systems are more miniaturized and less sophisticated than with software code. That makes embedded chips the most likely to fail, as opposed to simply making mistakes. "The condition that causes things to lock up, to freeze, is the unexpected logic condition."

And what could be more unexpected than 1999 turning into 1900? In some cases, date-sensitive embedded chips may indeed carry out their instructions, to everyone's chagrin. Take elevators with embedded "maintenance chips" that "learn" a century has suddenly passed. "It is very likely that several elevators will look at their dates, say 'it's been a long time since I had maintenance,' go down to the first floor, and stop functioning," Rafael Rodriguez, OIT's director of information systems architecture, told the Duke Y2K conference. Embedded chips may be inaccessible, but they are located in all kinds of devices that keep society running. A short list would include traffic lights, elevators, air-conditioning systems, medical devices, automobiles, and electric-power plants, not to forget computers themselves.

"The growth industry of the problem is embedded chips," says John Koskinen, a former deputy director for management for the U.S. Office of Management and Budget who early in 1998 became a presidential assistant and chair of the President's Council on the Year 2000 Conversion. "Over the last year or so, people have really focused on the fact that we run oil refineries, power plants, waste-treatment plants, and manufacturing plants all over the world with people sitting at computers responding to data that come from embedded sensors," he says. Estimates are that only 1 or 2 percent have date sensitivity. "But we shipped at least 5 billion chips in 1996, and 2 percent of 5 billion would be 100 million chips at significant risk."

As an OMB official, Koskinen supervised the federal government shutdowns of 1995-96. (He was also chair of Duke's board of trustees.) In a previous stint in the private sector, he helped reorganize large troubled firms like the Penn Central Transportation Company and Mutual Benefit Life Insurance Company. "I like crisis management," he says of his new job. "And it's hard to turn down the president and vice president, especially when it is a really national challenge." His assignment is to lead the charge to keep critical federal programs free from Y2K disruptions. He is also supposed to work with state and local governments, and to cooperate with private entities responsible for the nation's financial, telecommunications, public health, transportation, and electric-power generating systems.

Interviewed in June, the new appointee was already caught in political crossfire. "Under Koskinen, government performance has thundered from a D minus to an F," proclaimed Steve Horn, a Republican California congressman. Koskinen's own assessment is more sanguine. "The major systems of the United States will probably work all right," he says. "That includes the air-traffic control system, which has restructured the way they are managing the issue. Their backup system, which now works and is now compliant, will support about 70 percent of normal air traffic."

Running down the list of federal programs, Koskinen says that "Social Security, because they started making fixes in 1989, will not have a problem in terms of issuing their checks. They are still working on the disability program because that's administered by the states. And there are still states that are struggling to get their systems compliant. The Internal Revenue Service has major challenges. The Medicare/Medicaid processing system has major challenges."

Earlier, prognosticators had estimated that it would cost a whopping $30 billion to fix the federal government's Y2K problem. But the latest estimate is "closer to $5 billion," he says. As for the private sector, "I think in general the banking system will be in pretty good shape. There may be some small financial institutions that don't make it, but there are small financial institutions that don't make it every day." More ominously, "There are about 23 million small businesses out there, and a recent survey noted that 40 percent have no intention of doing anything about this problem," he says.

Another potential flash point is the securities market, which depends on computerized systems to move trading information and funds from country to country. Maggie Parent, a senior manager in the technology department of the Morgan Stanley global investment firm, is equivocal about Y2K's potential effects on her highly automated and increasingly internationalized industry. "The core infrastructure of the market, those kinds of entities which make up the backbone or central repositories for the marketplace, are by and large in very, very good shape," she says. "I'd say that's true in this country and, with some qualifications, it's true internationally. Where the challenge comes in is in the market participants, which have a tremendous diversity and complexity."

Stressing that her expertise is on the technical side rather than in investments, Parent notes that the international markets are linked by varieties of electronic systems and software programs, and that could result in logjams if some parts were disrupted. "For example, Country X is off-line, so investors who have money there can't get their money out," she says. "The assets are still there, but they just can't get to them. They can't move them. They can't sell them. If a stock market is down, or up only intermittently, you might not be able to make your trade happen. If foreign investors who normally buy a lot of U.S. government bonds were unable to participate in one of our auctions, that will throw the normal distribution of debt out of kilter. So it's possible that people will end up with assets tied up in different locations. Or their customers may not show up on the days that they usually buy."

On the positive side, major U.S. investment firms have already staged a simulated test of Wall Street's ability to communicate and trade on the new millennium's first business days--essentially by moving ahead the clock. As of late July, there were "no major glitches," The New York Times reported. "A lot of responsible activity has been happening on Wall Street for two or three years, so I think we're really in good shape," says Parent, who was heavily involved in that exercise. But she acknowledges a widespread concern about the readiness of most of Europe, now devoting the majority of its information-technology resources to the January 1999 conversion to a common currency. Likewise, much of Asia is preoccupied with financial problems and has neither the funds nor the focus for retooling. And reports from Latin America aren't encouraging either.

"I personally think there is no way this is going to be a non-event," says Parent. "At the best it will be a couple of days of disruptions, with little electronic blips and hiccups and some paper backlogs. There is a possibility that will happen, but I don't think that's the most likely outcome. Personally, I think it is likely we will have as much as a month of rickety market operations, with trouble spots in particular countries. We just need to keep working at it over the next seventeen months."

Some of the most pessimistic views on market repercussions from Y2K have come from Edward Yardini, chief economist of the global investment banking firm Deutsche Morgan Grenfell. Testifying before a November 1997 Senate subcommittee, Yardini predicted there was a 40 percent risk the Millennium Bug could provoke a worldwide recession lasting at least twelve months--one as severe as the 1973-74 global recession. Speaking more recently, Yardini has moved up that risk level to 70 percent.

"One of the good things that's happened is that people who are listened to closely have been strongly broadcasting the message that this is serious," says David Shumate, director of finance and administration for the Duke Management Company (DUMAC), which oversees university investments. DUMAC functions as a "manager of managers" who do the actual banking and investing of Duke assets, Shumate explains. So, as the "first tier of issues," his office is now busy making sure the "custodial bank"--Bankers Trust--is fixing its own millennium bugs. "The second tier is the portfolio managers. We're working with our managers now to make sure they're coming to terms with their own internal systems."

Especially with all the publicity, Shumate says he doesn't think that Y2K will catch investors unprepared. "They're not going to be asleep at the wheel on this. The investment managers assess all kinds of risks, and this is just one more--albeit a new one--that they have to assess in looking at the viability of a company."

Electric power utilities are also potentially prone to Y2K problems because the flow of power is monitored by built-in hardware and software clocks. Since most utilities are linked to one another in a grid-like fashion, glitches in one place could spread to others in a domino effect. Also keep in mind that New Year's Day is in midwinter. If the lights go out, the furnace might not work either. Reports from the front lines are mixed. None of the ten largest U.S. utilities had completed Y2K contingency plans, according to a June Senate panel survey.

But the word from Duke Energy Corporation, the utility that supplies the university with power, is more upbeat. "We started on it in 1996; the target is to have it all corrected by June of next year," says company spokesman Randy Wheeless. Not only is Duke Energy assessing and correcting its faulty software and hardware, the company is also sharing details on its fixes with neighboring utilities, including some trade secrets, he adds. "Because utilities are so interconnected, we may be perfectly fine, but if something happens at Carolina Power & Light or at South Carolina Electric, that could affect us, too." Like other utilities, Duke Energy also plans to have extra personnel on hand who can throw switches by hand if automated systems fail. And, because January 1, 2000, will be a major holiday, energy consumption will be lower than normal, even if a cold snap were to hit the region, Wheeless contends.

While the Internet is now alive with Y2K news stories, rumors, and warnings, planning and repairs have been quietly under way in some places, including Duke, for years. "I've been working on Year 2000 for about a decade now," says the medical center's David Kirby. "About half the work in our core systems has gotten done as a consequence of normal maintenance. And, of course, that's a lot easier than correcting a whole lot of things later and hoping that they work all right." Fixing programming problems means "touching code," the software equivalent of performing major surgery. Programmers must carefully open up the heart of a computer system's source codes, where code words are housed that direct the machine to do its job. Like a surgeon, a programmer must plan his or her operation well, starting with choosing which software "module" to enter.

"Most of the labor is tied up in selecting the module, re-analyzing what it does, and making sure you understand it, making and testing the change, then replacement of the change in the production environment," Kirby says. "So it's very efficient to make these date-related changes while you have to be in the module doing something else anyway." He says a combination of consultants and regular employees on overtime are performing this work. Overall, he "guesstimates" that correcting medical center Y2K bugs could cost between $4 million and $6 million overall. And he thinks it can get its "core" units--the largely mainframe-based systems that handle billings, appointments, orders, and inpatient and outpatient information--ready for Y2K in time to reserve the final year for testing.

But Kirby sees the core systems as only the hub of a wheel made up of other concentric rings. He thinks the systems most likely to have problems are in the outermost ring, composed of several hundred office networks that each link together individual personal computers. These smaller "local area networks" will be on their own to identify and correct their millennium bugs. "They may not be technically adept, and have no routine source of funding to do this kind of thing," he says.

Meanwhile, the medical center's clinical engineering department is systematically looking for faulty embedded hardware in its inventory of about 20,000 clinical devices. Even if its failure would not threaten lives, a Y2K-noncompliant device may still have to be junked. For example, a manufacturer of older defibrillators--devices that "shock" patients' aberrant heartbeats into stability--may not guarantee their product can accurately "time stamp" a record of a defibrillation incident that happens in the year 2000. That's unacceptable, because both doctors and lawyers rely on the accuracy of such records. "So here are $6,000 pieces of equipment that will have to be thrown away because they can't correctly create a record, even if they still technically work," says Kirby, who also focuses on how problems outside the campus will affect Duke. "There is probably nothing much we can do if the traffic lights are not going to work, but we have to be aware. And I expect someone at Duke to go to the city of Durham and ask: 'Is your payroll system Year 2000-ready?' And if the answer is anything other than crystal-clear, we might begin to hedge our bets about the bus system. If payroll doesn't pay, then the bus drivers aren't going to drive."

He also wonders about how well phone systems are apt to work in some more outlying clinics that interact with Duke, or about the e-mail connections. "Don't worry," he wryly tells people who ask. "Everything will not be all right."

Y2K is not just a coming event. Pager customers and broadcasters already got a taste of the future last May when the Galaxy IV communications satellite reportedly went out of control during a Year 2000-related test. Some of Duke's student information software, part of the core "Legacy" administrative systems maintained by OIT on the university's non-medical side, began operating in the twenty-first century back in 1996 in order to track properly the Class of 2000. The other Legacy systems that process university finances and employee payrolls and benefits are being "remediated" and will be fully tested in advance of the year 2000, says Nancy Wooters, the assistant director of OIT Application Services who is supervising the effort.

Programmers on both the medical and non-medical campuses are also making limited alterations to the remaining old code using a technique known as "windowing." Wooters explains that "windowing is not a permanent fix. You really haven't changed anything. You've just wrapped a protective shell around the code that is going to fail." Windowing lets programmers sidestep the problem of trying to change all 30 percent of the Duke software estimated to need conversion from two- to four-digit dates. Instead, it superimposes new logic instructions that will be good for many more years. For example, in the software that keeps track of students, "If you see a date that says '03, it's most likely not going to be 1903, but 2003," Wooters says. "So you can assume it's in the new century and can then make accurate decisions. I've found that, in industry, windowing is the preferred method now because of the lack of time."

Duke has already held a Year 2000 dress rehearsal for some of its remediated Legacy system code. The tests took place in October 1997, and again last April, at the Boulder, Colorado, "disaster recovery" facility where Duke's administrative software would be processed in the event of a hardware failure on campus. "What we did was answer some very basic questions, such as what happens to the mainframe on December 31, 1999," Wooters says. "Will it know what to do? Will it come up if we turn it off and then turn it back on January 1? No one knew. We had to find out." Moving the disaster recovery computer's clock ahead to 11:30 p.m. at the end of 1999, "We watched it roll over to 2000. It knew what date it was, and the date was correct."

The same thing happened when they shut the computer off in the old millennium and turned it back on in the new. But when programmers tried to make the computer process data, they found much of the critical software had been set to "expire" before 2000. They were able to overcome those problems, however, and began testing the payroll, financial, and student systems. "Screens came up," Wooters reports. "Batch jobs ran, and gave us good results. We printed payroll checks, and they were fine." She estimates revamping the Legacy core systems may cost as much as $1.2 million.

A nagging question remains. Why did a method for dealing with memory space shortages during the early days of computing persist over the next three decades, despite vast improvements in technology and sharp declines in its price? "It was a combination of forces," says the White House's Koskinen. "Once you start a two-digit identification as the standard, then every new system gets tied to the old system. There was also the continuing assumption that software systems were being upgraded and replaced so frequently that they weren't going to necessarily be there by the end of the Nineties."

The medical center's Kirby suggests "low motive" was also involved. "It's not like people in the software industry were standing around with nothing to do. We've always been busy. It's always been a fifty-hour-a-week kind of environment. There have always been things that we've regretted not being able to do just because there was not enough of a combination of resources and people time."

Noffsinger, the Vanstar consultant, says that computer memory actually only got cheap very recently. Four years ago, it cost him $800 to double his home computer capacity to 240 million bytes, "which I thought was just fantastic," he says. This summer, he paid $200 to upgrade an equivalent system to 5 billion bytes.

Then again, "we also got in the habit of doing it that way," he acknowledges. "We got in the habit."