Conduit, V4 N1, Spring 1995
Computers in the CS Museum - by Tom Doeppner
The faculty members who in 1979 became the CS department had been involved with computers at Brown for well over a decade. Andy van Dam's early work with undergraduates on hypertext, producing first the Hypertext Editing System and then FRESS, ran on the university's IBM mainframe-a 360 Model 67 owned by the university for an incredibly long time (from the late '60s till the late '70s). Our first significant computer system was BUGS, the Brown University Graphics System, and our first general-purpose computer was a VAX-11/780 (named Nancy when "the VAX" no longer sufficed to identify it) that arrived in early '79. Generalpurpose time sharing prospered here with the addition of another VAX-11/780 in 1982 (Sluggo) and finally a 12-processor Encore Multimax in 1987 (Zaphod). Another VAX, a 750 (Skyler), was acquired in '85 and used as our mail server. Nancy now rests in the department's computer museum.
Brown was a pioneer in using workstations for CS education as well as research. Our first Apollo workstation (model DN400) arrived in '81. By that fall we possessed seventeen DN400s and used them in the classroom for our introductory programming and algorithms courses. The DN400s were supplemented with the next-generation Apollo, the DN300, in mid- '83: sixty were acquired for instruction, fifteen for research. The instructional machines arrived in the first two weeks of the '83 fall semester, just after Jeff Coady, newly hired to administer them. Jeff, who had never seen a DN400 before, soon had to cope with running what was perhaps the largest collection of Apollos outside of Apollo. The Apollos were joined by a Sun 1 workstation in late '82 (Fritzi now rests in the computer museum). We acquired a couple of Sun 2s in mid-'84 (one of which, Munin, is now in the computer museum) and some Sun 3s in another couple of years. Our first Sun 4 arrived in mid-'87. Our original workstations weren't all that exciting to our AI folks. They investigated various Lisp packages for both the VAX and the Apollos, but finally decided that they would be best off with Symbolics Lisp Machines. Fortunately, money was found for these and five were acquired in '85 and '86-Babar, Bimbo, Clyde, Dumbo, and Horton. They served us well and were retired in '90: one of them now rests in the computer museum, the others were donated to Brown's Division of Engineering.
Late in the spring of 1988 we moved to our present quarters, in the newly constructed CIT building. We had hoped to install our recently ordered SPARCstation 1s in time for the fall semester, but instead, Sun leased us a number of Sun 3s and we used Zaphod (the Encore Multimax), originally a research machine, as our central facility. We chose not to continue using the VAXes, but sold Sluggo and Skyler; Nancy we kept (at least its primary cabinet- it had grown over the years into two large cabinets holding an impressive 10MB of primary storage, three good-sized disk drives, and two tape drives. All these latter items were disposed of). Nancy became the basis of our computer museum (at a time when VAX-11/780s were still being used at a number of other places). The SPARCstation 1s finally started to arrive in late winter of '89. By the summer we had enough of them that we had no further use for Zaphod, which we sold to Dick Bulterman, late of Brown and then (and now) of CWI in Amsterdam. By '92 our SPARCstation 1s had become a bit dated, so we replaced them with SPARCstation 10s, which now form the bulk of our computer holdings.
The Early Days
The initial configuration of BUGS, built by van Dam's graphics group, became operational in mid-'71. It consisted of a pair of Digital Scien tific Meta4 processors and a Vector General vector-graphics display and was augmented with Simale (Super-Integral Microprogrammable Arithmetic and Logic Expediter) in '75. Simale, designed and built by former undergraduate Harold Webber, had a four-processor 18-bit SIMD architecture-each processor had a 38-nanosecond cycle time for an effective peak performance of 105 MIPS. It supported real-time 3D and 4D vector graphics with matrix transformations, clipping, and dynamic level-of-detail management, and it was distinguished by never having a hardware failure in its seven-year lifetime-it was taken down only to replace light bulbs. Simale currently rests in the department's computer museum. BUGS was originally installed in the University Computer Center at 180 George St. It was connected to the 360/67 via RPC used for dynamic division of labor experiments between the mainframe host and the graphics satellite. We believe this was the first published use of RPC. When we moved to the new building (Kassar House at 151 Thayer Street) in May '79, BUGS moved to the basement, along with Nancy. Its tenth birthday was celebrated in the summer of '81 (those dealing with the VAX were explicitly not invited). It was decommissioned in early '82 when it became clear that there was no future in vector graphics and when work on the extension to 151 Thayer Street (Gould Lab) made part of the basement unusable.
UNIX Comes to Brown
In 1977 DEC announced the VAX-11/780. It was clear to us that this would be an ideal machine on which to run UNIX (an obscure research OS at the time); and it was also clear to us that the Program in CS (not yet a department) needed its own time-sharing system. NSF's new equipment program for CS departments granted us a bit over $100K to purchase our time-sharing system. This was about $100K less than we needed, but, with DEC's help, we were able to buy a VAX-11/780, configured with 512K of memory, one 67MB disk drive, and an amazingly slow tape drive. We considered purchasing a Prime 750 but, fortunately as it turned out, we stuck with our plans to get a VAX.
We intended to receive the VAX in fall '78. However, though renovation of our building at 151 Thayer Street had begun, it was in no shape to house a computer. We knew that there was "plenty" of space in the Barus-Holley and Prince Lab buildings-all we had to do was to get someone to part with some (temporarily as we hastened to point out). This turned out not to be easy. John Savage (acting director of the program at the time) and I had numerous conversations with our colleagues in Engineering and Physics. The room we thought was lined up fell through at the last minute (I placed a panic call to the loading dock of the DEC VAX factory one morning and convinced them not to ship our computer as it was about to be put on a truck.) By December we finally got a room (in Prince Lab) and the VAX arrived on January 8, 1979. UNIX was not quite ready for the VAX at this time (Bob Sedgewick, Steve Reiss, and I visited Bell Labs in summer '78 to check on its progress and were assured that it would be ready by early '79). So we ran VMS release 0.9, which came with no compilers and nothing of interest except for Adventure and a Scrabble game. (The student we hired to administer the system, Eric Albert, was a champion Scrabble player and enjoyed the game immensely.) In desperate need of a compiler, we became a beta site for DEC's Pascal compiler.
In May '79 the renovation of Kassar House was complete and the people of the department moved in, along with BUGS and the VAX. The building had been wired with RS232 cables and we were ready to put terminals in all faculty and student offices. Except we only owned four terminals. We had ordered four of DEC's new VT100s, but they were in short supply and we were allocated one; the others were due to arrive "soon." Even had they been readily available, they were too expensive for us to acquire in large numbers. So we settled for a cheaper alternative, the Zenith Z19 (also known as the Heathkit H19; despite our poverty, we did not acquire any in kit form). In the meantime, we had gone back to NSF and were awarded additional money. We used this to purchase another 67MB disk drive and 512K more memory. DEC helped out by granting us five VT100s, all of which arrived at the beginning of the '79 fall semester. One of these now rests in our computer museum. In June '79 UNIX finally arrived. It was release 32V which, as advertised, ran on the VAX, but did not support virtual memory. But it had a compiler and all sorts of other nifty tools and the department finally entered the computer age. (We held no grudges and invited the graphics people to Nancy's fifth birthday party in '84.) Once we had a C compiler, Steve Reiss wrote b, the first generation of the Brown editor. This was quickly adopted by most of the department and its successor, bb, is still used by a few diehards (including me). With UNIX came email. Initially it was only for use in the department-we had no network connections. But in fall '79 I established our first email link to another computer-to "research" at Bell Labs via uucp, the Unix-to- Unix Copy program. This was a poor-man's approach to networks-point-to-point connections via phone lines. But it worked and eventually gave us world-wide (if slow and unreliable) email connections. In '82 we joined CSNET and had substantially improved mail service and in '86 we became connected to the Internet.
Late in '79 Berkeley UNIX was introduced and we were one of the first recipients of 3BSD, the first version of UNIX to support virtual memory. It was notable for being impressively slow-it compared unfavorably with VMS in many benchmarks, but even so, few people wanted to run VMS. Less than a year later we installed 4BSD, a much faster version of Berkeley UNIX, followed by 4.1BSD, which we installed as soon as it was available. In '82, as part of research collaboration with DEC, we were granted another VAX-11/780 (Sluggo) for support of graphics work. This was installed on October 6, 1982, just in time for the dedication of Gould Lab that evening. It now became important to get into the UNIX networking business, so we bought two ethernet boards and some cable and acquired an experimental version of Berkeley UNIX, 4.1aBSD (installed in November '82), which added networking support to 4.1BSD.
In August '83 we installed 4.1cBSD, which fixed a number of longstanding problems with UNIX, such as its file system. This was the biggest change since we started with UNIX. Despite a number of warnings, many people were caught off guard and had to do a lot of last-minute scrambling to get their code working again. However, the switch to the next official release, 4.2BSD, was made in late fall and hardly anyone noticed. 4.3BSD was introduced a few years later and again no one noticed.
Graphics, IBM, and Construction
With the demise of BUGS, the graphics group entered the worlds of raster graphics and UNIX. A special graphics room was built in the basement of Kassar and called "BURGER"-Brown University Raster Graphics Experimentation Room (constructed by graduate student and master carpenter Bill Smith). In it was installed in December '79 our first (color) raster display, a Ramtek 9400, currently in our computer museum. A number of notable software projects used this display, including the Interactive Graphical Documents project, BRUWIN (the Brown University Window Manager), and the 4D animation project. Eventually the Ramtek was joined by a couple of Lexidata raster-graphics displays.
In '80 and '81 we began collaborative work with IBM. We needed IBM hardware for this work, so we began thinking about where to put an IBM computer. Serious thought was given to installing the computer in the Kassar House garage (despite the objections of those of us who parked our bicycles there). We eventually decided that the garage would go away to make room for Gould Lab, so the acquisition of IBM hardware was postponed.
When construction of Gould Lab began, a number of changes had to be made to let the construction workers use portions of the basement. BUGS was demolished and Nancy was moved into its place. (By this time Nancy had innumerable terminal connections, etc., so moving it was no easy chore.) BURGER became the construction crew's office, so a small corner room of the basement was taken over for the graphics lab and christened "microBURGER." Running the computers while construction was going on was interesting. Amazing amounts of dust were kicked up, so the computer areas of the basement were sealed off with plastic sheets. When construction work was particularly heavy, the machines had to be taken down so that the disk drives wouldn't be damaged by the vibrations. Circuit boards were frequently reseated.
We were still thinking about an IBM installation and suddenly realized that it would put major demands on the basement air conditioning. We had to up the requirements for the air conditioner, which produced a considerable increase in the size of the air conditioner-so much so that the air conditioner required would not fit through any of the openings into the basement. So the construction crew removed a number of stones from the basement walls to make an opening (just) big enough to put the air conditioner through. We hired riggers (Zavota Brothers) to slide the air conditioner through the hole and set it up in the basement. This was spectacular to watch. They brought in some impressive equipment and some incredibly strong people and got the job done in seemingly no time at all, without a scratch to either building or air conditioner.
Finally, pretty much at the last minute, everything was cleaned up in time for Gould Lab's dedication on October 6, 1982. There was now room in the basement, so an IBM 4381 was installed and the department had its first (and so far, only) IBM mainframe, which was used for research on text processing. It was removed a year or so later.
Graphics moved out of the basement and into a spiffy lab within Gould Lab. BURGER was history. The Ramtek and Lexidatas were moved in and were joined by high-end Apollos and a top-of-the-line Evans and Sutherland PS-300 vector display (which now rests in the computer museum).
We became intrigued with the idea of workstations in the late '70s when we heard about what was going on at Xerox PARC. Finally in 1980 Three Rivers Computer announced the Perq workstation (but didn't deliver it until much later). This at least made it clear that workstations were about to become commercially available. One of the things that we wanted to use workstations for was instruction, so we applied to NSF's CAUSE program (Comprehensive Assistance to Undergraduate Science Education-a program that got the axe under the Reagan administration and were awarded $150K. Workstations back then were being quoted for ~$35K each (they weren't being delivered yet), so this was not a whole lot of money. But it was something and we started searching for a workstation vendor.
It was clear to us that we had to make ourselves look exciting so that we could get some assistance (i.e., attractive discounts) from vendors. We put together a brochure describing our needs and our vision for instructional computing. We commissioned an artist to draw a picture of our proposed lab for the brochure and we designed Gould Lab to feature a computerized classroom (holding up to sixty workstations), which became known as the Foxboro Auditorium. We sent the brochure to a number of prospective vendors and donors. We narrowed things down to three serious potential vendors: Three Rivers, Xerox, and Apollo. Three Rivers was the early favorite, since they had actually announced a commercial product. Xerox, unlike anyone else, had actually produced workstations. Apollo was run by people who were already successful in the computer business (a number had come from Prime and had unsuccessfully attempted to sell us a Prime 750 a few years earlier). We had pretty well decided upon Apollo, but then new developments occurred at Xerox, so we delayed our order. This cost us the honor of receiving the first Apollos shipped. Things became clearer at Xerox in a few days so we put in a firm order to Apollo for seventeen workstations, two with disk drives (33MB each). We received the fifth and sixth machines shipped, in March '81. One of these, "node C", now rests in our computer museum (it was retired from active duty in May 1988). The other machines trickled in and students began to use them in the '81/'82 academic year. One of the more popular first applications written was "PACman", a copy of a then-popular computer game.
We became big proponents of workstations and Apollos, giving numerous demos for Apollo's potential customers. Marc Brown, a grad student/ staff member, founded the Apollo Users Group and organized its first meeting, held at the Biltmore Hotel in downtown Providence in the summer of '82. Great progress was made in instructional software and in Bob Sedgewick's and Marc Brown's research on algorithm animation (resulting in BALSA). This pioneering work helped us immensely in obtaining our first major equipment grant-$3 million over five years from NSF's CER (coordinated experimental research) program, awarded in May 1982.
We used this grant to purchase additional Apollo workstations, used for research, as well as additional memory and disk drives, etc., for Nancy (the first VAX). The first generation of Apollo workstation was rather bulky. When the Foxboro auditorium was ready in September '82, though we only had seventeen machines, they occupied a fair amount of space. The next generation of Apollo workstation, the DN300, was considerably smaller . With the help of a generous grant from the Exxon Education Foundation, we were able to purchase enough DN300s to populate the Foxboro Auditorium. As mentioned earlier, these arrived all at once at the beginning of the fall semester in '83. One of these machines, "Node 87C", remains, resting in the computer museum. Rather than the two file servers serving fifteen diskless machines in our first-generation classroom, we now had sixty diskless instructional machines (along with fifteen research machines, each with a 33MB disk) served by sixteen file servers. We finally had the instructional facility we had described in our brochure three-and-a-half years earlier.
The Apollos lasted us for the remainder of our days in Kassar House and Gould Lab. But by 1987 it was clear that our Apollos were rapidly becoming obsolete. So we embarked on another round of choosing a workstation vendor. We talked to a number of vendors and things finally worked out into a contest between DEC, NeXT, and Sun (Apollo dropped out, since they didn't have a machine that met our needs at our price). DEC proposed a VAX-based workstation. This seemed to be a safe choice-it would run the same operating system we were running on the big VAXes. However, the workstation VAX was pretty slow compared to the machine Sun was proposing. (Little did we know that DEC had just started a project to produce a MIPS-based workstation. They couldn't tell us about this until several months later, but even so, we would never have believed that it would be shipping in nine months.) NeXT seemed pretty exciting, and we were the subject of a very impressive sales pitch by Steve Jobs. Sun was proposing some exciting hardware, but they didn't have it ready to demonstrate for us. After a lot of discussion and a few benchmarks, we decided in April '88 to go with Sun.
None of the old Apollos made the move to the new CIT building, except for the two that went to the museum. Unfortunately the Sun SPARCstation 1s weren't ready until March '89, so the new teaching lab, christened the "Sun Lab", was filled with Sun3s, leased to us through Sun. But by summer '89 we had a full complement of SPARCstation 1s.
Our most recent round of vendor selection started in '91 and finished in early '92 by choosing Sun again. Our current SPARCstation 10s are aging rapidly and will soon be considerably slower than the PCs owned by many of our undergraduates. We have a recapitalization plan in place; the next chapter in the workstation story will begin to unfold within the next couple of years.