Stanford University Home

Stanford News Archive

Stanford Report, July 9, 2003

Campus digital workhorse running final laps after 35 years

BY BARBARA PALMER

The reports of its demise have been exaggerated over the years, but this time it's true, say administrators in Information Technology Systems and Services (ITSS). The mainframe, the university's digital workhorse for decades, is being put out to pasture.

Between now and Aug. 31, owners of mainframe accounts need to look at files and data stored on the mainframe and move to alternative technologies anything they need to keep, said Regina Speer, campus readiness manager for the ITSS mainframe retirement team.

On Sept. 1 -- not coincidentally the same date that the new web-based Oracle financial system is scheduled to go live for campus administrators -- accounts that have not been identified as necessary to perform vital business functions or scheduled to be switched over later will be closed. The mainframe is set to retire on Dec. 31. (More information about how to get off the mainframe is available at http://mainframe-retirement.stanford.edu.)

Meighan McWilliam, technology account manager and part of the mainframe retirement project team, now spends a majority of her workdays -- and some of her weekends -- helping departments close down the 6,000 or so accounts that still reside on the mainframe. At its peak, there were 30,000 accounts, including noncampus users, on the mainframe. Some of what McWilliam is doing requires real detective work, since many of the people who opened the accounts have long ago left Stanford or retired.

 

John Vier
Photo: L.A. Cicero

A diagram of the Forsythe Data Center that hangs outside the second-floor office of John Vier, ITSS manager of facility services, graphically illustrates the way distributed computing systems have taken over the mainframes. On the diagram, the mainframes are represented by four little rectangles -- two peach, one aqua and one lime green -- surrounded on all sides by a mosaic representing the hundreds of servers that now carry most of Stanford's information processing load. At one time, said Vier as he walked amid the data center's humming servers, the mainframe dominated the 16,000-square-foot area.

Dick Guertin
Photo: L.A. Cicero

The remarkable thing is not that the mainframe will be retired -- the 35-year-old system is "like a jet slowly losing its engines," said Dick Guertin, a software developer who has worked at Stanford since 1970. What is amazing, say those whose work lives have spanned much of the computer revolution, is that the mainframe system and the ground-breaking applications designed here to run on it have held up so well for so long.

Jane Marcus
Photo: L.A. Cicero

"WYLBUR is still in my fingers," said Jane Marcus, an academic computing consultant and member of the mainframe retirement team, referring to a 1967 program that allowed users to create text documents on terminals connected to the mainframe.

Marcus, a former English teacher who came to Stanford to study the influence of television on education, was sitting at a mainframe-connected terminal in 1977 when revelation struck. "I thought, 'Oh, this is going to change everything,'" she said. "'Computers are going to be much more powerful for education than television.'"

And Stanford was squarely at the forefront of innovation. Programs like ORVYL, MILTEN and WYLBUR, which created ways for users to share files and access the mainframe's computational power, were revolutionary at the time, she said.

Christine Moe
Photo: L.A. Cicero

"It was a time when working with computers was intoxicating," said Christine Moe, a systems programmer who came to Stanford in 1977 and worked on the project that first hooked the mainframe up to a network using file transfer protocols and ASCII terminals.

Guertin and software developer programmer Bill Kiefer arrived even earlier. Both came to Stanford within weeks of each other in 1970 and worked together on SPIRES, the Stanford Public Information Retrieval System. Funded by a National Science Foundation grant, SPIRES was a second iteration of the Stanford Physics Information Retrieval System, also called SPIRES, which gave high-energy physicists a way to research documents online. The generalized version of SPIRES was intended as a software tool that could be used to help automate the university's libraries and a wide variety of research and record-keeping functions. In 1971, SPIRES ran on an IBM 360/67 with 1 megabyte of memory and allowed 90 typewriter terminals to be online at one time.

Two decades later, SPIRES provided the first "killer app" for the World Wide Web, when its high-energy physics database was searched from Switzerland. "I'm not sure that most of the people who work here at the university know that SPIRES got started here serving the high-energy physics community, and that's how the web got started at CERN [European Laboratory for Particle Physics] in Switzerland," said Marcus. "It was the very same academic community that started SPIRES here -- and it changed the world."

Bill Kiefer
Photo: L.A. Cicero

The genius of programmers like Guertin and Kiefer was their ability to find solutions to specific problems that could be generalized across the university, said John Klemm, who came to Stanford in 1978 and wrote hundreds of pages of documentation for SPIRES.

'Seat of the pants' programming

In the early 1980s, programmers developed an array of administrative systems using SPIRES: NSI (Network for Student Information), which automated student records for staff use; PRISM, an online collection of administrative files; and SUFIN, the Stanford University Financial Information Network. As a result, Stanford's major administrative systems were automated long before any other university's, Marcus said.

"We did it all by the seat of our pants," said Kiefer. When requests for different functions would be made, "We'd say, 'Oh, let's see if we can make it happen.'

"One thing about Stanford that is so different from industry is that about every one of these projects started from particular people who said, 'Oh, I know how to do this,'" he continued. "Bill Witscher, for example, helped design the NSI system. He just loved to do it. He had a knack for it. And he had a knack for taking one or two people with good ideas and to put these things together. They never went out and made a mint on these things, but they did wonderful work."

"This was a time when the entrepreneurial mindset was here at Stanford -- and not just the ones that got away, like Sun and Cisco [both founded by former Stanford faculty, staff and students], but for the people who were here," said Moe. Managers had "an uncanny knack of recognizing talent and running the roadblocks. For years, we were left alone."

"Dick and I were in a little room ... designing the next version of SPIRES and [Systems Group Manager] Jim Moore would come in once a year and say, 'I don't know what you're doing, but keep it up,'" Kiefer said.

John Klemm
Photo: L.A. Cicero

Through the 1980s, "any night you came here you would find anywhere from 10 to 20 people working until midnight -- plus dealing with the 10 or 20 or 30 people who were connected from home," Klemm said. The glamour has faded, he said. "There isn't the excitement anymore of working on something that you're making for your friends from scratch."

The mainframe environment was "a completely separate world the way the Internet is a different world now to people," Klemm said. "There were all sorts of things going on simultaneously. People writing their theses on the mainframe; people entering administrative data, writing purchase orders; people playing games, trading restaurant reviews, recipes and jokes."

And decades before it proliferated widely, e-mail. Employees sent instant messages and a form of e-mail in the late 1970s, making Stanford one of the very few companies or institutions with electronic mail at the time, Moe said.

"The mainframe allowed you to have a community," Marcus said. And then "just as we were getting this group of people together who understood the power of e-mail, the PC came along," she said. The personal computer changed the computing paradigm, and it would take years before the networks linking PCs could provide a way to communicate electronically that matched what had long been available to users of the mainframe, she added.

'Instant gratification'

Given the ability to distribute computer processing power over networks and the growth in sophistication of commercial software systems, it no longer makes sense for the university to custom-build mainframe computer systems -- and many of the engineers and programmers with the skills to do it have left or are reaching retirement age. Kiefer said he may retire next May when his age and years of service equal a nice round number: 100.

"Sometimes I almost feel guilty that we built this for the university," said Kiefer. The tailor-made administrative systems addressed user needs so specifically that it has been difficult to replace the homegrown applications with off-the-shelf commercial software, he said. But given the evolution of computer systems, "that's the way it has to be done," he said.

"What SPIRES gave our community was almost instant gratification," Guertin said. "If [someone] would come to us and say, 'We need August to have 32, 33, 34' -- dates beyond the end of the month -- two days later we could call them on the phone and say, 'Try this system, it's got what you want.' But if you called a vendor and said, 'We need this capability in your system,' the answer would be 'maybe next year' or maybe 'never.'"

"What's pretty fascinating about SPIRES is that the databases built in the early 1970s are still around and we've never had to rebuild most of them," Kiefer said. And although most applications that ran on the mainframe can be migrated to other technologies, there are some that can't be replaced, he said. Guertin has written a program that emulates SPIRES on a UNIX platform. "The mainframe is retiring, but I wouldn't say that SPIRES is retiring," Guertin said.

Jim Nisbet used to ride his bike to campus while he was still a Menlo Park high school student to crash computer classes being held in Pine Hall. He'd stay around afterward to root in the trash so he "could look at what people typed and learn," he said. Nisbet was asked to leave -- and was hired to work on mainframe systems at SLAC in 1972, he said.

Nisbet, who founded and sold a database software company after leaving Stanford in 1991, said his perspective is similar to that of someone who was around for the earliest days of aviation. "You can watch this incredible change take place, from something that could just barely take off to something that is going at supersonic speed. It's kind of neat to see the Stanford mainframe survived all these transitions."

In the computer industry, which measures years like dog years, it's a real feat that systems built for the mainframe lasted so long, he said. "If we could get 10 or 15 years out of [new systems], that's something that people would kill for."

Mainframe computers
Photo: L.A. Cicero